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, int oif,
  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         fl6.flowi6_oif = 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
 113         return 0;
 114 }
 115
 116 static inline void
 117 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 118 {
 119         struct flowi6 *fl6 = &fl->u.ip6;
 120         int onlyproto = 0;
 121         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 122         u16 offset = sizeof(*hdr);
 123         struct ipv6_opt_hdr *exthdr;
 124         const unsigned char *nh = skb_network_header(skb);
 125         u16 nhoff = IP6CB(skb)->nhoff;
 126         int oif = 0;
 127         u8 nexthdr;
 128
 129         if (!nhoff)
 130                 nhoff = offsetof(struct ipv6hdr, nexthdr);
 131
 132         nexthdr = nh[nhoff];
 133
 134         if (skb_dst(skb))
 135                 oif = skb_dst(skb)->dev->ifindex;
 136
 137         memset(fl6, 0, sizeof(struct flowi6));
 138         fl6->flowi6_mark = skb->mark;
 139         fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 140
 141         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 142         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 143
 144         while (nh + offset + 1 < skb->data ||
 145                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 146                 nh = skb_network_header(skb);
 147                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 148
 149                 switch (nexthdr) {
 150                 case NEXTHDR_FRAGMENT:
 151                         onlyproto = 1;
 152                 case NEXTHDR_ROUTING:
 153                 case NEXTHDR_HOP:
 154                 case NEXTHDR_DEST:
 155                         offset += ipv6_optlen(exthdr);
 156                         nexthdr = exthdr->nexthdr;
 157                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 158                         break;
 159
 160                 case IPPROTO_UDP:
 161                 case IPPROTO_UDPLITE:
 162                 case IPPROTO_TCP:
 163                 case IPPROTO_SCTP:
 164                 case IPPROTO_DCCP:
 165                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 166                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 167                                 __be16 *ports;
 168
 169                                 nh = skb_network_header(skb);
 170                                 ports = (__be16 *)(nh + offset);
 171                                 fl6->fl6_sport = ports[!!reverse];
 172                                 fl6->fl6_dport = ports[!reverse];
 173                         }
 174                         fl6->flowi6_proto = nexthdr;
 175                         return;
 176
 177                 case IPPROTO_ICMPV6:
 178                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 179                                 u8 *icmp;
 180
 181                                 nh = skb_network_header(skb);
 182                                 icmp = (u8 *)(nh + offset);
 183                                 fl6->fl6_icmp_type = icmp[0];
 184                                 fl6->fl6_icmp_code = icmp[1];
 185                         }
 186                         fl6->flowi6_proto = nexthdr;
 187                         return;
 188
 189 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 190                 case IPPROTO_MH:
 191                         offset += ipv6_optlen(exthdr);
 192                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 193                                 struct ip6_mh *mh;
 194
 195                                 nh = skb_network_header(skb);
 196                                 mh = (struct ip6_mh *)(nh + offset);
 197                                 fl6->fl6_mh_type = mh->ip6mh_type;
 198                         }
 199                         fl6->flowi6_proto = nexthdr;
 200                         return;
 201 #endif
 202
 203                 /* XXX Why are there these headers? */
 204                 case IPPROTO_AH:
 205                 case IPPROTO_ESP:
 206                 case IPPROTO_COMP:
 207                 default:
 208                         fl6->fl6_ipsec_spi = 0;
 209                         fl6->flowi6_proto = nexthdr;
 210                         return;
 211                 }
 212         }
 213 }
 214
 215 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 216 {
 217         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 218
 219         xfrm6_policy_afinfo.garbage_collect(net);
 220         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 221 }
 222
 223 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 224                               struct sk_buff *skb, u32 mtu)
 225 {
 226         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 227         struct dst_entry *path = xdst->route;
 228
 229         path->ops->update_pmtu(path, sk, skb, mtu);
 230 }
 231
 232 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 233                            struct sk_buff *skb)
 234 {
 235         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 236         struct dst_entry *path = xdst->route;
 237
 238         path->ops->redirect(path, sk, skb);
 239 }
 240
 241 static void xfrm6_dst_destroy(struct dst_entry *dst)
 242 {
 243         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 244
 245         if (likely(xdst->u.rt6.rt6i_idev))
 246                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 247         dst_destroy_metrics_generic(dst);
 248         xfrm_dst_destroy(xdst);
 249 }
 250
 251 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 252                              int unregister)
 253 {
 254         struct xfrm_dst *xdst;
 255
 256         if (!unregister)
 257                 return;
 258
 259         xdst = (struct xfrm_dst *)dst;
 260         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 261                 struct inet6_dev *loopback_idev =
 262                         in6_dev_get(dev_net(dev)->loopback_dev);
 263                 BUG_ON(!loopback_idev);
 264
 265                 do {
 266                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 267                         xdst->u.rt6.rt6i_idev = loopback_idev;
 268                         in6_dev_hold(loopback_idev);
 269                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 270                 } while (xdst->u.dst.xfrm);
 271
 272                 __in6_dev_put(loopback_idev);
 273         }
 274
 275         xfrm_dst_ifdown(dst, dev);
 276 }
 277
 278 static struct dst_ops xfrm6_dst_ops = {
 279         .family =               AF_INET6,
 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_path =            xfrm6_init_path,
 298         .fill_dst =             xfrm6_fill_dst,
 299         .blackhole_route =      ip6_blackhole_route,
 300 };
 301
 302 static int __init xfrm6_policy_init(void)
 303 {
 304         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 305 }
 306
 307 static void xfrm6_policy_fini(void)
 308 {
 309         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 310 }
 311
 312 #ifdef CONFIG_SYSCTL
 313 static struct ctl_table xfrm6_policy_table[] = {
 314         {
 315                 .procname       = "xfrm6_gc_thresh",
 316                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 317                 .maxlen         = sizeof(int),
 318                 .mode           = 0644,
 319                 .proc_handler   = proc_dointvec,
 320         },
 321         { }
 322 };
 323
 324 static int __net_init xfrm6_net_init(struct net *net)
 325 {
 326         struct ctl_table *table;
 327         struct ctl_table_header *hdr;
 328
 329         table = xfrm6_policy_table;
 330         if (!net_eq(net, &init_net)) {
 331                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 332                 if (!table)
 333                         goto err_alloc;
 334
 335                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 336         }
 337
 338         hdr = register_net_sysctl(net, "net/ipv6", table);
 339         if (!hdr)
 340                 goto err_reg;
 341
 342         net->ipv6.sysctl.xfrm6_hdr = hdr;
 343         return 0;
 344
 345 err_reg:
 346         if (!net_eq(net, &init_net))
 347                 kfree(table);
 348 err_alloc:
 349         return -ENOMEM;
 350 }
 351
 352 static void __net_exit xfrm6_net_exit(struct net *net)
 353 {
 354         struct ctl_table *table;
 355
 356         if (!net->ipv6.sysctl.xfrm6_hdr)
 357                 return;
 358
 359         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 360         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 361         if (!net_eq(net, &init_net))
 362                 kfree(table);
 363 }
 364
 365 static struct pernet_operations xfrm6_net_ops = {
 366         .init   = xfrm6_net_init,
 367         .exit   = xfrm6_net_exit,
 368 };
 369 #endif
 370
 371 int __init xfrm6_init(void)
 372 {
 373         int ret;
 374
 375         dst_entries_init(&xfrm6_dst_ops);
 376
 377         ret = xfrm6_policy_init();
 378         if (ret) {
 379                 dst_entries_destroy(&xfrm6_dst_ops);
 380                 goto out;
 381         }
 382         ret = xfrm6_state_init();
 383         if (ret)
 384                 goto out_policy;
 385
 386         ret = xfrm6_protocol_init();
 387         if (ret)
 388                 goto out_state;
 389
 390 #ifdef CONFIG_SYSCTL
 391         register_pernet_subsys(&xfrm6_net_ops);
 392 #endif
 393 out:
 394         return ret;
 395 out_state:
 396         xfrm6_state_fini();
 397 out_policy:
 398         xfrm6_policy_fini();
 399         goto out;
 400 }
 401
 402 void xfrm6_fini(void)
 403 {
 404 #ifdef CONFIG_SYSCTL
 405         unregister_pernet_subsys(&xfrm6_net_ops);
 406 #endif
 407         xfrm6_protocol_fini();
 408         xfrm6_policy_fini();
 409         xfrm6_state_fini();
 410         dst_entries_destroy(&xfrm6_dst_ops);
 411 }