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 #include <net/l3mdev.h>
  24 #if IS_ENABLED(CONFIG_IPV6_MIP6)
  25 #include <net/mip6.h>
  26 #endif
  27
  28 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, int oif,
  29                                           const xfrm_address_t *saddr,
  30                                           const xfrm_address_t *daddr)
  31 {
  32         struct flowi6 fl6;
  33         struct dst_entry *dst;
  34         int err;
  35
  36         memset(&fl6, 0, sizeof(fl6));
  37         fl6.flowi6_oif = l3mdev_master_ifindex_by_index(net, oif);
  38         fl6.flowi6_flags = FLOWI_FLAG_SKIP_NH_OIF;
  39         memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
  40         if (saddr)
  41                 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
  42
  43         dst = ip6_route_output(net, NULL, &fl6);
  44
  45         err = dst->error;
  46         if (dst->error) {
  47                 dst_release(dst);
  48                 dst = ERR_PTR(err);
  49         }
  50
  51         return dst;
  52 }
  53
  54 static int xfrm6_get_saddr(struct net *net, int oif,
  55                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  56 {
  57         struct dst_entry *dst;
  58         struct net_device *dev;
  59
  60         dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr);
  61         if (IS_ERR(dst))
  62                 return -EHOSTUNREACH;
  63
  64         dev = ip6_dst_idev(dst)->dev;
  65         ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
  66         dst_release(dst);
  67         return 0;
  68 }
  69
  70 static int xfrm6_get_tos(const struct flowi *fl)
  71 {
  72         return 0;
  73 }
  74
  75 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  76                            int nfheader_len)
  77 {
  78         if (dst->ops->family == AF_INET6) {
  79                 struct rt6_info *rt = (struct rt6_info *)dst;
  80                 path->path_cookie = rt6_get_cookie(rt);
  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         xdst->route_cookie = rt6_get_cookie(rt);
 109         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 110         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 111         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 112         INIT_LIST_HEAD(&xdst->u.rt6.rt6i_uncached);
 113         rt6_uncached_list_add(&xdst->u.rt6);
 114
 115         return 0;
 116 }
 117
 118 static inline void
 119 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 120 {
 121         struct flowi6 *fl6 = &fl->u.ip6;
 122         int onlyproto = 0;
 123         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 124         u16 offset = sizeof(*hdr);
 125         struct ipv6_opt_hdr *exthdr;
 126         const unsigned char *nh = skb_network_header(skb);
 127         u16 nhoff = IP6CB(skb)->nhoff;
 128         int oif = 0;
 129         u8 nexthdr;
 130
 131         if (!nhoff)
 132                 nhoff = offsetof(struct ipv6hdr, nexthdr);
 133
 134         nexthdr = nh[nhoff];
 135
 136         if (skb_dst(skb))
 137                 oif = skb_dst(skb)->dev->ifindex;
 138
 139         memset(fl6, 0, sizeof(struct flowi6));
 140         fl6->flowi6_mark = skb->mark;
 141         fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 142
 143         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 144         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 145
 146         while (nh + offset + 1 < skb->data ||
 147                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 148                 nh = skb_network_header(skb);
 149                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 150
 151                 switch (nexthdr) {
 152                 case NEXTHDR_FRAGMENT:
 153                         onlyproto = 1;
 154                 case NEXTHDR_ROUTING:
 155                 case NEXTHDR_HOP:
 156                 case NEXTHDR_DEST:
 157                         offset += ipv6_optlen(exthdr);
 158                         nexthdr = exthdr->nexthdr;
 159                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 160                         break;
 161
 162                 case IPPROTO_UDP:
 163                 case IPPROTO_UDPLITE:
 164                 case IPPROTO_TCP:
 165                 case IPPROTO_SCTP:
 166                 case IPPROTO_DCCP:
 167                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 168                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 169                                 __be16 *ports;
 170
 171                                 nh = skb_network_header(skb);
 172                                 ports = (__be16 *)(nh + offset);
 173                                 fl6->fl6_sport = ports[!!reverse];
 174                                 fl6->fl6_dport = ports[!reverse];
 175                         }
 176                         fl6->flowi6_proto = nexthdr;
 177                         return;
 178
 179                 case IPPROTO_ICMPV6:
 180                         if (!onlyproto && (nh + offset + 2 < skb->data ||
 181                             pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
 182                                 u8 *icmp;
 183
 184                                 nh = skb_network_header(skb);
 185                                 icmp = (u8 *)(nh + offset);
 186                                 fl6->fl6_icmp_type = icmp[0];
 187                                 fl6->fl6_icmp_code = icmp[1];
 188                         }
 189                         fl6->flowi6_proto = nexthdr;
 190                         return;
 191
 192 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 193                 case IPPROTO_MH:
 194                         offset += ipv6_optlen(exthdr);
 195                         if (!onlyproto && (nh + offset + 3 < skb->data ||
 196                             pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
 197                                 struct ip6_mh *mh;
 198
 199                                 nh = skb_network_header(skb);
 200                                 mh = (struct ip6_mh *)(nh + offset);
 201                                 fl6->fl6_mh_type = mh->ip6mh_type;
 202                         }
 203                         fl6->flowi6_proto = nexthdr;
 204                         return;
 205 #endif
 206
 207                 /* XXX Why are there these headers? */
 208                 case IPPROTO_AH:
 209                 case IPPROTO_ESP:
 210                 case IPPROTO_COMP:
 211                 default:
 212                         fl6->fl6_ipsec_spi = 0;
 213                         fl6->flowi6_proto = nexthdr;
 214                         return;
 215                 }
 216         }
 217 }
 218
 219 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 220 {
 221         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 222
 223         xfrm_garbage_collect_deferred(net);
 224         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 225 }
 226
 227 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 228                               struct sk_buff *skb, u32 mtu)
 229 {
 230         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 231         struct dst_entry *path = xdst->route;
 232
 233         path->ops->update_pmtu(path, sk, skb, mtu);
 234 }
 235
 236 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 237                            struct sk_buff *skb)
 238 {
 239         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 240         struct dst_entry *path = xdst->route;
 241
 242         path->ops->redirect(path, sk, skb);
 243 }
 244
 245 static void xfrm6_dst_destroy(struct dst_entry *dst)
 246 {
 247         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 248
 249         if (likely(xdst->u.rt6.rt6i_idev))
 250                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 251         dst_destroy_metrics_generic(dst);
 252         if (xdst->u.rt6.rt6i_uncached_list)
 253                 rt6_uncached_list_del(&xdst->u.rt6);
 254         xfrm_dst_destroy(xdst);
 255 }
 256
 257 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 258                              int unregister)
 259 {
 260         struct xfrm_dst *xdst;
 261
 262         if (!unregister)
 263                 return;
 264
 265         xdst = (struct xfrm_dst *)dst;
 266         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 267                 struct inet6_dev *loopback_idev =
 268                         in6_dev_get(dev_net(dev)->loopback_dev);
 269                 BUG_ON(!loopback_idev);
 270
 271                 do {
 272                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 273                         xdst->u.rt6.rt6i_idev = loopback_idev;
 274                         in6_dev_hold(loopback_idev);
 275                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 276                 } while (xdst->u.dst.xfrm);
 277
 278                 __in6_dev_put(loopback_idev);
 279         }
 280
 281         xfrm_dst_ifdown(dst, dev);
 282 }
 283
 284 static struct dst_ops xfrm6_dst_ops_template = {
 285         .family =               AF_INET6,
 286         .gc =                   xfrm6_garbage_collect,
 287         .update_pmtu =          xfrm6_update_pmtu,
 288         .redirect =             xfrm6_redirect,
 289         .cow_metrics =          dst_cow_metrics_generic,
 290         .destroy =              xfrm6_dst_destroy,
 291         .ifdown =               xfrm6_dst_ifdown,
 292         .local_out =            __ip6_local_out,
 293         .gc_thresh =            INT_MAX,
 294 };
 295
 296 static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 297         .dst_ops =              &xfrm6_dst_ops_template,
 298         .dst_lookup =           xfrm6_dst_lookup,
 299         .get_saddr =            xfrm6_get_saddr,
 300         .decode_session =       _decode_session6,
 301         .get_tos =              xfrm6_get_tos,
 302         .init_path =            xfrm6_init_path,
 303         .fill_dst =             xfrm6_fill_dst,
 304         .blackhole_route =      ip6_blackhole_route,
 305 };
 306
 307 static int __init xfrm6_policy_init(void)
 308 {
 309         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
 310 }
 311
 312 static void xfrm6_policy_fini(void)
 313 {
 314         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 315 }
 316
 317 #ifdef CONFIG_SYSCTL
 318 static struct ctl_table xfrm6_policy_table[] = {
 319         {
 320                 .procname       = "xfrm6_gc_thresh",
 321                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 322                 .maxlen         = sizeof(int),
 323                 .mode           = 0644,
 324                 .proc_handler   = proc_dointvec,
 325         },
 326         { }
 327 };
 328
 329 static int __net_init xfrm6_net_sysctl_init(struct net *net)
 330 {
 331         struct ctl_table *table;
 332         struct ctl_table_header *hdr;
 333
 334         table = xfrm6_policy_table;
 335         if (!net_eq(net, &init_net)) {
 336                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 337                 if (!table)
 338                         goto err_alloc;
 339
 340                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 341         }
 342
 343         hdr = register_net_sysctl(net, "net/ipv6", table);
 344         if (!hdr)
 345                 goto err_reg;
 346
 347         net->ipv6.sysctl.xfrm6_hdr = hdr;
 348         return 0;
 349
 350 err_reg:
 351         if (!net_eq(net, &init_net))
 352                 kfree(table);
 353 err_alloc:
 354         return -ENOMEM;
 355 }
 356
 357 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
 358 {
 359         struct ctl_table *table;
 360
 361         if (!net->ipv6.sysctl.xfrm6_hdr)
 362                 return;
 363
 364         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 365         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 366         if (!net_eq(net, &init_net))
 367                 kfree(table);
 368 }
 369 #else /* CONFIG_SYSCTL */
 370 static inline int xfrm6_net_sysctl_init(struct net *net)
 371 {
 372         return 0;
 373 }
 374
 375 static inline void xfrm6_net_sysctl_exit(struct net *net)
 376 {
 377 }
 378 #endif
 379
 380 static int __net_init xfrm6_net_init(struct net *net)
 381 {
 382         int ret;
 383
 384         memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
 385                sizeof(xfrm6_dst_ops_template));
 386         ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
 387         if (ret)
 388                 return ret;
 389
 390         ret = xfrm6_net_sysctl_init(net);
 391         if (ret)
 392                 dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
 393
 394         return ret;
 395 }
 396
 397 static void __net_exit xfrm6_net_exit(struct net *net)
 398 {
 399         xfrm6_net_sysctl_exit(net);
 400         dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
 401 }
 402
 403 static struct pernet_operations xfrm6_net_ops = {
 404         .init   = xfrm6_net_init,
 405         .exit   = xfrm6_net_exit,
 406 };
 407
 408 int __init xfrm6_init(void)
 409 {
 410         int ret;
 411
 412         ret = xfrm6_policy_init();
 413         if (ret)
 414                 goto out;
 415         ret = xfrm6_state_init();
 416         if (ret)
 417                 goto out_policy;
 418
 419         ret = xfrm6_protocol_init();
 420         if (ret)
 421                 goto out_state;
 422
 423         register_pernet_subsys(&xfrm6_net_ops);
 424 out:
 425         return ret;
 426 out_state:
 427         xfrm6_state_fini();
 428 out_policy:
 429         xfrm6_policy_fini();
 430         goto out;
 431 }
 432
 433 void xfrm6_fini(void)
 434 {
 435         unregister_pernet_subsys(&xfrm6_net_ops);
 436         xfrm6_protocol_fini();
 437         xfrm6_policy_fini();
 438         xfrm6_state_fini();
 439 }