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         u16 offset = skb_network_header_len(skb);
 134         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 135         struct ipv6_opt_hdr *exthdr;
 136         const unsigned char *nh = skb_network_header(skb);
 137         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 138         int oif = 0;
 139
 140         if (skb_dst(skb))
 141                 oif = skb_dst(skb)->dev->ifindex;
 142
 143         memset(fl6, 0, sizeof(struct flowi6));
 144         fl6->flowi6_mark = skb->mark;
 145         fl6->flowi6_oif = reverse ? skb->skb_iif : oif;
 146
 147         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 148         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 149
 150         while (nh + offset + 1 < skb->data ||
 151                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 152                 nh = skb_network_header(skb);
 153                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 154
 155                 switch (nexthdr) {
 156                 case NEXTHDR_FRAGMENT:
 157                         onlyproto = 1;
 158                 case NEXTHDR_ROUTING:
 159                 case NEXTHDR_HOP:
 160                 case NEXTHDR_DEST:
 161                         offset += ipv6_optlen(exthdr);
 162                         nexthdr = exthdr->nexthdr;
 163                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 164                         break;
 165
 166                 case IPPROTO_UDP:
 167                 case IPPROTO_UDPLITE:
 168                 case IPPROTO_TCP:
 169                 case IPPROTO_SCTP:
 170                 case IPPROTO_DCCP:
 171                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 172                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 173                                 __be16 *ports = (__be16 *)exthdr;
 174
 175                                 fl6->fl6_sport = ports[!!reverse];
 176                                 fl6->fl6_dport = ports[!reverse];
 177                         }
 178                         fl6->flowi6_proto = nexthdr;
 179                         return;
 180
 181                 case IPPROTO_ICMPV6:
 182                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 183                                 u8 *icmp = (u8 *)exthdr;
 184
 185                                 fl6->fl6_icmp_type = icmp[0];
 186                                 fl6->fl6_icmp_code = icmp[1];
 187                         }
 188                         fl6->flowi6_proto = nexthdr;
 189                         return;
 190
 191 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 192                 case IPPROTO_MH:
 193                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 194                                 struct ip6_mh *mh;
 195                                 mh = (struct ip6_mh *)exthdr;
 196
 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         if (rt6_has_peer(&xdst->u.rt6)) {
 249                 struct inet_peer *peer = rt6_peer_ptr(&xdst->u.rt6);
 250                 inet_putpeer(peer);
 251         }
 252         xfrm_dst_destroy(xdst);
 253 }
 254
 255 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 256                              int unregister)
 257 {
 258         struct xfrm_dst *xdst;
 259
 260         if (!unregister)
 261                 return;
 262
 263         xdst = (struct xfrm_dst *)dst;
 264         if (xdst->u.rt6.rt6i_idev->dev == dev) {
 265                 struct inet6_dev *loopback_idev =
 266                         in6_dev_get(dev_net(dev)->loopback_dev);
 267                 BUG_ON(!loopback_idev);
 268
 269                 do {
 270                         in6_dev_put(xdst->u.rt6.rt6i_idev);
 271                         xdst->u.rt6.rt6i_idev = loopback_idev;
 272                         in6_dev_hold(loopback_idev);
 273                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 274                 } while (xdst->u.dst.xfrm);
 275
 276                 __in6_dev_put(loopback_idev);
 277         }
 278
 279         xfrm_dst_ifdown(dst, dev);
 280 }
 281
 282 static struct dst_ops xfrm6_dst_ops = {
 283         .family =               AF_INET6,
 284         .protocol =             cpu_to_be16(ETH_P_IPV6),
 285         .gc =                   xfrm6_garbage_collect,
 286         .update_pmtu =          xfrm6_update_pmtu,
 287         .redirect =             xfrm6_redirect,
 288         .cow_metrics =          dst_cow_metrics_generic,
 289         .destroy =              xfrm6_dst_destroy,
 290         .ifdown =               xfrm6_dst_ifdown,
 291         .local_out =            __ip6_local_out,
 292         .gc_thresh =            32768,
 293 };
 294
 295 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 296         .family =               AF_INET6,
 297         .dst_ops =              &xfrm6_dst_ops,
 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_dst =             xfrm6_init_dst,
 303         .init_path =            xfrm6_init_path,
 304         .fill_dst =             xfrm6_fill_dst,
 305         .blackhole_route =      ip6_blackhole_route,
 306 };
 307
 308 static int __init xfrm6_policy_init(void)
 309 {
 310         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 311 }
 312
 313 static void xfrm6_policy_fini(void)
 314 {
 315         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 316 }
 317
 318 #ifdef CONFIG_SYSCTL
 319 static struct ctl_table xfrm6_policy_table[] = {
 320         {
 321                 .procname       = "xfrm6_gc_thresh",
 322                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 323                 .maxlen         = sizeof(int),
 324                 .mode           = 0644,
 325                 .proc_handler   = proc_dointvec,
 326         },
 327         { }
 328 };
 329
 330 static int __net_init xfrm6_net_init(struct net *net)
 331 {
 332         struct ctl_table *table;
 333         struct ctl_table_header *hdr;
 334
 335         table = xfrm6_policy_table;
 336         if (!net_eq(net, &init_net)) {
 337                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 338                 if (!table)
 339                         goto err_alloc;
 340
 341                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 342         }
 343
 344         hdr = register_net_sysctl(net, "net/ipv6", table);
 345         if (!hdr)
 346                 goto err_reg;
 347
 348         net->ipv6.sysctl.xfrm6_hdr = hdr;
 349         return 0;
 350
 351 err_reg:
 352         if (!net_eq(net, &init_net))
 353                 kfree(table);
 354 err_alloc:
 355         return -ENOMEM;
 356 }
 357
 358 static void __net_exit xfrm6_net_exit(struct net *net)
 359 {
 360         struct ctl_table *table;
 361
 362         if (net->ipv6.sysctl.xfrm6_hdr == NULL)
 363                 return;
 364
 365         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 366         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 367         if (!net_eq(net, &init_net))
 368                 kfree(table);
 369 }
 370
 371 static struct pernet_operations xfrm6_net_ops = {
 372         .init   = xfrm6_net_init,
 373         .exit   = xfrm6_net_exit,
 374 };
 375 #endif
 376
 377 int __init xfrm6_init(void)
 378 {
 379         int ret;
 380
 381         dst_entries_init(&xfrm6_dst_ops);
 382
 383         ret = xfrm6_policy_init();
 384         if (ret) {
 385                 dst_entries_destroy(&xfrm6_dst_ops);
 386                 goto out;
 387         }
 388         ret = xfrm6_state_init();
 389         if (ret)
 390                 goto out_policy;
 391
 392 #ifdef CONFIG_SYSCTL
 393         register_pernet_subsys(&xfrm6_net_ops);
 394 #endif
 395 out:
 396         return ret;
 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_policy_fini();
 408         xfrm6_state_fini();
 409         dst_entries_destroy(&xfrm6_dst_ops);
 410 }