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                                           u32 mark)
  32 {
  33         struct flowi6 fl6;
  34         struct dst_entry *dst;
  35         int err;
  36
  37         memset(&fl6, 0, sizeof(fl6));
  38         fl6.flowi6_oif = l3mdev_master_ifindex_by_index(net, oif);
  39         fl6.flowi6_flags = FLOWI_FLAG_SKIP_NH_OIF;
  40         fl6.flowi6_mark = mark;
  41         memcpy(&fl6.daddr, daddr, sizeof(fl6.daddr));
  42         if (saddr)
  43                 memcpy(&fl6.saddr, saddr, sizeof(fl6.saddr));
  44
  45         dst = ip6_route_output(net, NULL, &fl6);
  46
  47         err = dst->error;
  48         if (dst->error) {
  49                 dst_release(dst);
  50                 dst = ERR_PTR(err);
  51         }
  52
  53         return dst;
  54 }
  55
  56 static int xfrm6_get_saddr(struct net *net, int oif,
  57                            xfrm_address_t *saddr, xfrm_address_t *daddr,
  58                            u32 mark)
  59 {
  60         struct dst_entry *dst;
  61         struct net_device *dev;
  62
  63         dst = xfrm6_dst_lookup(net, 0, oif, NULL, daddr, mark);
  64         if (IS_ERR(dst))
  65                 return -EHOSTUNREACH;
  66
  67         dev = ip6_dst_idev(dst)->dev;
  68         ipv6_dev_get_saddr(dev_net(dev), dev, &daddr->in6, 0, &saddr->in6);
  69         dst_release(dst);
  70         return 0;
  71 }
  72
  73 static int xfrm6_get_tos(const struct flowi *fl)
  74 {
  75         return 0;
  76 }
  77
  78 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  79                            int nfheader_len)
  80 {
  81         if (dst->ops->family == AF_INET6) {
  82                 struct rt6_info *rt = (struct rt6_info *)dst;
  83                 path->path_cookie = rt6_get_cookie(rt);
  84         }
  85
  86         path->u.rt6.rt6i_nfheader_len = nfheader_len;
  87
  88         return 0;
  89 }
  90
  91 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  92                           const struct flowi *fl)
  93 {
  94         struct rt6_info *rt = (struct rt6_info *)xdst->route;
  95
  96         xdst->u.dst.dev = dev;
  97         dev_hold(dev);
  98
  99         xdst->u.rt6.rt6i_idev = in6_dev_get(dev);
 100         if (!xdst->u.rt6.rt6i_idev) {
 101                 dev_put(dev);
 102                 return -ENODEV;
 103         }
 104
 105         /* Sheit... I remember I did this right. Apparently,
 106          * it was magically lost, so this code needs audit */
 107         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 108                                                    RTF_LOCAL);
 109         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 110         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 111         xdst->route_cookie = rt6_get_cookie(rt);
 112         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 113         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 114         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 115
 116         return 0;
 117 }
 118
 119 static inline void
 120 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 121 {
 122         struct flowi6 *fl6 = &fl->u.ip6;
 123         int onlyproto = 0;
 124         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 125         u16 offset = sizeof(*hdr);
 126         struct ipv6_opt_hdr *exthdr;
 127         const unsigned char *nh = skb_network_header(skb);
 128         u16 nhoff = IP6CB(skb)->nhoff;
 129         int oif = 0;
 130         u8 nexthdr;
 131
 132         if (!nhoff)
 133                 nhoff = offsetof(struct ipv6hdr, nexthdr);
 134
 135         nexthdr = nh[nhoff];
 136
 137         if (skb_dst(skb))
 138                 oif = skb_dst(skb)->dev->ifindex;
 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 && (nh + offset + 2 < skb->data ||
 182                             pskb_may_pull(skb, nh + offset + 2 - skb->data))) {
 183                                 u8 *icmp;
 184
 185                                 nh = skb_network_header(skb);
 186                                 icmp = (u8 *)(nh + offset);
 187                                 fl6->fl6_icmp_type = icmp[0];
 188                                 fl6->fl6_icmp_code = icmp[1];
 189                         }
 190                         fl6->flowi6_proto = nexthdr;
 191                         return;
 192
 193 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 194                 case IPPROTO_MH:
 195                         offset += ipv6_optlen(exthdr);
 196                         if (!onlyproto && (nh + offset + 3 < skb->data ||
 197                             pskb_may_pull(skb, nh + offset + 3 - skb->data))) {
 198                                 struct ip6_mh *mh;
 199
 200                                 nh = skb_network_header(skb);
 201                                 mh = (struct ip6_mh *)(nh + offset);
 202                                 fl6->fl6_mh_type = mh->ip6mh_type;
 203                         }
 204                         fl6->flowi6_proto = nexthdr;
 205                         return;
 206 #endif
 207
 208                 /* XXX Why are there these headers? */
 209                 case IPPROTO_AH:
 210                 case IPPROTO_ESP:
 211                 case IPPROTO_COMP:
 212                 default:
 213                         fl6->fl6_ipsec_spi = 0;
 214                         fl6->flowi6_proto = nexthdr;
 215                         return;
 216                 }
 217         }
 218 }
 219
 220 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 221                               struct sk_buff *skb, u32 mtu)
 222 {
 223         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 224         struct dst_entry *path = xdst->route;
 225
 226         path->ops->update_pmtu(path, sk, skb, mtu);
 227 }
 228
 229 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 230                            struct sk_buff *skb)
 231 {
 232         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 233         struct dst_entry *path = xdst->route;
 234
 235         path->ops->redirect(path, sk, skb);
 236 }
 237
 238 static void xfrm6_dst_destroy(struct dst_entry *dst)
 239 {
 240         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 241
 242         if (likely(xdst->u.rt6.rt6i_idev))
 243                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 244         dst_destroy_metrics_generic(dst);
 245         xfrm_dst_destroy(xdst);
 246 }
 247
 248 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 249                              int unregister)
 250 {
 251         struct xfrm_dst *xdst;
 252
 253         if (!unregister)
 254                 return;
 255
 256         xdst = (struct xfrm_dst *)dst;
 257         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 258                 struct inet6_dev *loopback_idev =
 259                         in6_dev_get(dev_net(dev)->loopback_dev);
 260                 BUG_ON(!loopback_idev);
 261
 262                 do {
 263                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 264                         xdst->u.rt6.rt6i_idev = loopback_idev;
 265                         in6_dev_hold(loopback_idev);
 266                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 267                 } while (xdst->u.dst.xfrm);
 268
 269                 __in6_dev_put(loopback_idev);
 270         }
 271
 272         xfrm_dst_ifdown(dst, dev);
 273 }
 274
 275 static struct dst_ops xfrm6_dst_ops_template = {
 276         .family =               AF_INET6,
 277         .update_pmtu =          xfrm6_update_pmtu,
 278         .redirect =             xfrm6_redirect,
 279         .cow_metrics =          dst_cow_metrics_generic,
 280         .destroy =              xfrm6_dst_destroy,
 281         .ifdown =               xfrm6_dst_ifdown,
 282         .local_out =            __ip6_local_out,
 283         .gc_thresh =            32768,
 284 };
 285
 286 static const struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 287         .dst_ops =              &xfrm6_dst_ops_template,
 288         .dst_lookup =           xfrm6_dst_lookup,
 289         .get_saddr =            xfrm6_get_saddr,
 290         .decode_session =       _decode_session6,
 291         .get_tos =              xfrm6_get_tos,
 292         .init_path =            xfrm6_init_path,
 293         .fill_dst =             xfrm6_fill_dst,
 294         .blackhole_route =      ip6_blackhole_route,
 295 };
 296
 297 static int __init xfrm6_policy_init(void)
 298 {
 299         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo, AF_INET6);
 300 }
 301
 302 static void xfrm6_policy_fini(void)
 303 {
 304         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 305 }
 306
 307 #ifdef CONFIG_SYSCTL
 308 static struct ctl_table xfrm6_policy_table[] = {
 309         {
 310                 .procname       = "xfrm6_gc_thresh",
 311                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 312                 .maxlen         = sizeof(int),
 313                 .mode           = 0644,
 314                 .proc_handler   = proc_dointvec,
 315         },
 316         { }
 317 };
 318
 319 static int __net_init xfrm6_net_sysctl_init(struct net *net)
 320 {
 321         struct ctl_table *table;
 322         struct ctl_table_header *hdr;
 323
 324         table = xfrm6_policy_table;
 325         if (!net_eq(net, &init_net)) {
 326                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 327                 if (!table)
 328                         goto err_alloc;
 329
 330                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 331         }
 332
 333         hdr = register_net_sysctl(net, "net/ipv6", table);
 334         if (!hdr)
 335                 goto err_reg;
 336
 337         net->ipv6.sysctl.xfrm6_hdr = hdr;
 338         return 0;
 339
 340 err_reg:
 341         if (!net_eq(net, &init_net))
 342                 kfree(table);
 343 err_alloc:
 344         return -ENOMEM;
 345 }
 346
 347 static void __net_exit xfrm6_net_sysctl_exit(struct net *net)
 348 {
 349         struct ctl_table *table;
 350
 351         if (!net->ipv6.sysctl.xfrm6_hdr)
 352                 return;
 353
 354         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 355         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 356         if (!net_eq(net, &init_net))
 357                 kfree(table);
 358 }
 359 #else /* CONFIG_SYSCTL */
 360 static inline int xfrm6_net_sysctl_init(struct net *net)
 361 {
 362         return 0;
 363 }
 364
 365 static inline void xfrm6_net_sysctl_exit(struct net *net)
 366 {
 367 }
 368 #endif
 369
 370 static int __net_init xfrm6_net_init(struct net *net)
 371 {
 372         int ret;
 373
 374         memcpy(&net->xfrm.xfrm6_dst_ops, &xfrm6_dst_ops_template,
 375                sizeof(xfrm6_dst_ops_template));
 376         ret = dst_entries_init(&net->xfrm.xfrm6_dst_ops);
 377         if (ret)
 378                 return ret;
 379
 380         ret = xfrm6_net_sysctl_init(net);
 381         if (ret)
 382                 dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
 383
 384         return ret;
 385 }
 386
 387 static void __net_exit xfrm6_net_exit(struct net *net)
 388 {
 389         xfrm6_net_sysctl_exit(net);
 390         dst_entries_destroy(&net->xfrm.xfrm6_dst_ops);
 391 }
 392
 393 static struct pernet_operations xfrm6_net_ops = {
 394         .init   = xfrm6_net_init,
 395         .exit   = xfrm6_net_exit,
 396 };
 397
 398 int __init xfrm6_init(void)
 399 {
 400         int ret;
 401
 402         ret = xfrm6_policy_init();
 403         if (ret)
 404                 goto out;
 405         ret = xfrm6_state_init();
 406         if (ret)
 407                 goto out_policy;
 408
 409         ret = xfrm6_protocol_init();
 410         if (ret)
 411                 goto out_state;
 412
 413         register_pernet_subsys(&xfrm6_net_ops);
 414 out:
 415         return ret;
 416 out_state:
 417         xfrm6_state_fini();
 418 out_policy:
 419         xfrm6_policy_fini();
 420         goto out;
 421 }
 422
 423 void xfrm6_fini(void)
 424 {
 425         unregister_pernet_subsys(&xfrm6_net_ops);
 426         xfrm6_protocol_fini();
 427         xfrm6_policy_fini();
 428         xfrm6_state_fini();
 429 }