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/vrf.h>
  24 #if IS_ENABLED(CONFIG_IPV6_MIP6)
  25 #include <net/mip6.h>
  26 #endif
  27
  28 static struct xfrm_policy_afinfo xfrm6_policy_afinfo;
  29
  30 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, int oif,
  31                                           const xfrm_address_t *saddr,
  32                                           const xfrm_address_t *daddr)
  33 {
  34         struct flowi6 fl6;
  35         struct dst_entry *dst;
  36         int err;
  37
  38         memset(&fl6, 0, sizeof(fl6));
  39         fl6.flowi6_oif = oif;
  40         memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
  41         if (saddr)
  42                 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
  43
  44         dst = ip6_route_output(net, NULL, &fl6);
  45
  46         err = dst->error;
  47         if (dst->error) {
  48                 dst_release(dst);
  49                 dst = ERR_PTR(err);
  50         }
  51
  52         return dst;
  53 }
  54
  55 static int xfrm6_get_saddr(struct net *net, int oif,
  56                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  57 {
  58         struct dst_entry *dst;
  59         struct net_device *dev;
  60
  61         dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr);
  62         if (IS_ERR(dst))
  63                 return -EHOSTUNREACH;
  64
  65         dev = ip6_dst_idev(dst)->dev;
  66         ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
  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 int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  77                            int nfheader_len)
  78 {
  79         if (dst->ops->family == AF_INET6) {
  80                 struct rt6_info *rt = (struct rt6_info *)dst;
  81                 path->path_cookie = rt6_get_cookie(rt);
  82         }
  83
  84         path->u.rt6.rt6i_nfheader_len = nfheader_len;
  85
  86         return 0;
  87 }
  88
  89 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  90                           const struct flowi *fl)
  91 {
  92         struct rt6_info *rt = (struct rt6_info *)xdst->route;
  93
  94         xdst->u.dst.dev = dev;
  95         dev_hold(dev);
  96
  97         xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
  98         if (!xdst->u.rt6.rt6i_idev) {
  99                 dev_put(dev);
 100                 return -ENODEV;
 101         }
 102
 103         /* Sheit... I remember I did this right. Apparently,
 104          * it was magically lost, so this code needs audit */
 105         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 106                                                    RTF_LOCAL);
 107         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 108         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 109         xdst->route_cookie = rt6_get_cookie(rt);
 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 = vrf_master_ifindex(skb_dst(skb)->dev) ?
 137                         : skb_dst(skb)->dev->ifindex;
 138         }
 139
 140         memset(fl6, 0, sizeof(struct flowi6));
 141         fl6->flowi6_mark = skb->mark;
 142         fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 143
 144         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 145         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 146
 147         while (nh + offset + 1 < skb->data ||
 148                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 149                 nh = skb_network_header(skb);
 150                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 151
 152                 switch (nexthdr) {
 153                 case NEXTHDR_FRAGMENT:
 154                         onlyproto = 1;
 155                 case NEXTHDR_ROUTING:
 156                 case NEXTHDR_HOP:
 157                 case NEXTHDR_DEST:
 158                         offset += ipv6_optlen(exthdr);
 159                         nexthdr = exthdr->nexthdr;
 160                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 161                         break;
 162
 163                 case IPPROTO_UDP:
 164                 case IPPROTO_UDPLITE:
 165                 case IPPROTO_TCP:
 166                 case IPPROTO_SCTP:
 167                 case IPPROTO_DCCP:
 168                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 169                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 170                                 __be16 *ports;
 171
 172                                 nh = skb_network_header(skb);
 173                                 ports = (__be16 *)(nh + offset);
 174                                 fl6->fl6_sport = ports[!!reverse];
 175                                 fl6->fl6_dport = ports[!reverse];
 176                         }
 177                         fl6->flowi6_proto = nexthdr;
 178                         return;
 179
 180                 case IPPROTO_ICMPV6:
 181                         if (!onlyproto && 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 && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 196                                 struct ip6_mh *mh;
 197
 198                                 nh = skb_network_header(skb);
 199                                 mh = (struct ip6_mh *)(nh + offset);
 200                                 fl6->fl6_mh_type = mh->ip6mh_type;
 201                         }
 202                         fl6->flowi6_proto = nexthdr;
 203                         return;
 204 #endif
 205
 206                 /* XXX Why are there these headers? */
 207                 case IPPROTO_AH:
 208                 case IPPROTO_ESP:
 209                 case IPPROTO_COMP:
 210                 default:
 211                         fl6->fl6_ipsec_spi = 0;
 212                         fl6->flowi6_proto = nexthdr;
 213                         return;
 214                 }
 215         }
 216 }
 217
 218 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 219 {
 220         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 221
 222         xfrm6_policy_afinfo.garbage_collect(net);
 223         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 224 }
 225
 226 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 227                               struct sk_buff *skb, u32 mtu)
 228 {
 229         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 230         struct dst_entry *path = xdst->route;
 231
 232         path->ops->update_pmtu(path, sk, skb, mtu);
 233 }
 234
 235 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 236                            struct sk_buff *skb)
 237 {
 238         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 239         struct dst_entry *path = xdst->route;
 240
 241         path->ops->redirect(path, sk, skb);
 242 }
 243
 244 static void xfrm6_dst_destroy(struct dst_entry *dst)
 245 {
 246         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 247
 248         if (likely(xdst->u.rt6.rt6i_idev))
 249                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 250         dst_destroy_metrics_generic(dst);
 251         xfrm_dst_destroy(xdst);
 252 }
 253
 254 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 255                              int unregister)
 256 {
 257         struct xfrm_dst *xdst;
 258
 259         if (!unregister)
 260                 return;
 261
 262         xdst = (struct xfrm_dst *)dst;
 263         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 264                 struct inet6_dev *loopback_idev =
 265                         in6_dev_get(dev_net(dev)->loopback_dev);
 266                 BUG_ON(!loopback_idev);
 267
 268                 do {
 269                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 270                         xdst->u.rt6.rt6i_idev = loopback_idev;
 271                         in6_dev_hold(loopback_idev);
 272                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 273                 } while (xdst->u.dst.xfrm);
 274
 275                 __in6_dev_put(loopback_idev);
 276         }
 277
 278         xfrm_dst_ifdown(dst, dev);
 279 }
 280
 281 static struct dst_ops xfrm6_dst_ops = {
 282         .family =               AF_INET6,
 283         .gc =                   xfrm6_garbage_collect,
 284         .update_pmtu =          xfrm6_update_pmtu,
 285         .redirect =             xfrm6_redirect,
 286         .cow_metrics =          dst_cow_metrics_generic,
 287         .destroy =              xfrm6_dst_destroy,
 288         .ifdown =               xfrm6_dst_ifdown,
 289         .local_out =            __ip6_local_out,
 290         .gc_thresh =            32768,
 291 };
 292
 293 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 294         .family =               AF_INET6,
 295         .dst_ops =              &xfrm6_dst_ops,
 296         .dst_lookup =           xfrm6_dst_lookup,
 297         .get_saddr =            xfrm6_get_saddr,
 298         .decode_session =       _decode_session6,
 299         .get_tos =              xfrm6_get_tos,
 300         .init_path =            xfrm6_init_path,
 301         .fill_dst =             xfrm6_fill_dst,
 302         .blackhole_route =      ip6_blackhole_route,
 303 };
 304
 305 static int __init xfrm6_policy_init(void)
 306 {
 307         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 308 }
 309
 310 static void xfrm6_policy_fini(void)
 311 {
 312         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 313 }
 314
 315 #ifdef CONFIG_SYSCTL
 316 static struct ctl_table xfrm6_policy_table[] = {
 317         {
 318                 .procname       = "xfrm6_gc_thresh",
 319                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 320                 .maxlen         = sizeof(int),
 321                 .mode           = 0644,
 322                 .proc_handler   = proc_dointvec,
 323         },
 324         { }
 325 };
 326
 327 static int __net_init xfrm6_net_init(struct net *net)
 328 {
 329         struct ctl_table *table;
 330         struct ctl_table_header *hdr;
 331
 332         table = xfrm6_policy_table;
 333         if (!net_eq(net, &init_net)) {
 334                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 335                 if (!table)
 336                         goto err_alloc;
 337
 338                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 339         }
 340
 341         hdr = register_net_sysctl(net, "net/ipv6", table);
 342         if (!hdr)
 343                 goto err_reg;
 344
 345         net->ipv6.sysctl.xfrm6_hdr = hdr;
 346         return 0;
 347
 348 err_reg:
 349         if (!net_eq(net, &init_net))
 350                 kfree(table);
 351 err_alloc:
 352         return -ENOMEM;
 353 }
 354
 355 static void __net_exit xfrm6_net_exit(struct net *net)
 356 {
 357         struct ctl_table *table;
 358
 359         if (!net->ipv6.sysctl.xfrm6_hdr)
 360                 return;
 361
 362         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 363         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 364         if (!net_eq(net, &init_net))
 365                 kfree(table);
 366 }
 367
 368 static struct pernet_operations xfrm6_net_ops = {
 369         .init   = xfrm6_net_init,
 370         .exit   = xfrm6_net_exit,
 371 };
 372 #endif
 373
 374 int __init xfrm6_init(void)
 375 {
 376         int ret;
 377
 378         dst_entries_init(&xfrm6_dst_ops);
 379
 380         ret = xfrm6_policy_init();
 381         if (ret) {
 382                 dst_entries_destroy(&xfrm6_dst_ops);
 383                 goto out;
 384         }
 385         ret = xfrm6_state_init();
 386         if (ret)
 387                 goto out_policy;
 388
 389         ret = xfrm6_protocol_init();
 390         if (ret)
 391                 goto out_state;
 392
 393 #ifdef CONFIG_SYSCTL
 394         register_pernet_subsys(&xfrm6_net_ops);
 395 #endif
 396 out:
 397         return ret;
 398 out_state:
 399         xfrm6_state_fini();
 400 out_policy:
 401         xfrm6_policy_fini();
 402         goto out;
 403 }
 404
 405 void xfrm6_fini(void)
 406 {
 407 #ifdef CONFIG_SYSCTL
 408         unregister_pernet_subsys(&xfrm6_net_ops);
 409 #endif
 410         xfrm6_protocol_fini();
 411         xfrm6_policy_fini();
 412         xfrm6_state_fini();
 413         dst_entries_destroy(&xfrm6_dst_ops);
 414 }