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                 return -ENODEV;
 108
 109         rt6_transfer_peer(&xdst->u.rt6, rt);
 110
 111         /* Sheit... I remember I did this right. Apparently,
 112          * it was magically lost, so this code needs audit */
 113         xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST |
 114                                                    RTF_LOCAL);
 115         xdst->u.rt6.rt6i_metric = rt->rt6i_metric;
 116         xdst->u.rt6.rt6i_node = rt->rt6i_node;
 117         if (rt->rt6i_node)
 118                 xdst->route_cookie = rt->rt6i_node->fn_sernum;
 119         xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway;
 120         xdst->u.rt6.rt6i_dst = rt->rt6i_dst;
 121         xdst->u.rt6.rt6i_src = rt->rt6i_src;
 122
 123         return 0;
 124 }
 125
 126 static inline void
 127 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse)
 128 {
 129         struct flowi6 *fl6 = &fl->u.ip6;
 130         int onlyproto = 0;
 131         u16 offset = skb_network_header_len(skb);
 132         const struct ipv6hdr *hdr = ipv6_hdr(skb);
 133         struct ipv6_opt_hdr *exthdr;
 134         const unsigned char *nh = skb_network_header(skb);
 135         u8 nexthdr = nh[IP6CB(skb)->nhoff];
 136
 137         memset(fl6, 0, sizeof(struct flowi6));
 138         fl6->flowi6_mark = skb->mark;
 139
 140         fl6->daddr = reverse ? hdr->saddr : hdr->daddr;
 141         fl6->saddr = reverse ? hdr->daddr : hdr->saddr;
 142
 143         while (nh + offset + 1 < skb->data ||
 144                pskb_may_pull(skb, nh + offset + 1 - skb->data)) {
 145                 nh = skb_network_header(skb);
 146                 exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 147
 148                 switch (nexthdr) {
 149                 case NEXTHDR_FRAGMENT:
 150                         onlyproto = 1;
 151                 case NEXTHDR_ROUTING:
 152                 case NEXTHDR_HOP:
 153                 case NEXTHDR_DEST:
 154                         offset += ipv6_optlen(exthdr);
 155                         nexthdr = exthdr->nexthdr;
 156                         exthdr = (struct ipv6_opt_hdr *)(nh + offset);
 157                         break;
 158
 159                 case IPPROTO_UDP:
 160                 case IPPROTO_UDPLITE:
 161                 case IPPROTO_TCP:
 162                 case IPPROTO_SCTP:
 163                 case IPPROTO_DCCP:
 164                         if (!onlyproto && (nh + offset + 4 < skb->data ||
 165                              pskb_may_pull(skb, nh + offset + 4 - skb->data))) {
 166                                 __be16 *ports = (__be16 *)exthdr;
 167
 168                                 fl6->fl6_sport = ports[!!reverse];
 169                                 fl6->fl6_dport = ports[!reverse];
 170                         }
 171                         fl6->flowi6_proto = nexthdr;
 172                         return;
 173
 174                 case IPPROTO_ICMPV6:
 175                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) {
 176                                 u8 *icmp = (u8 *)exthdr;
 177
 178                                 fl6->fl6_icmp_type = icmp[0];
 179                                 fl6->fl6_icmp_code = icmp[1];
 180                         }
 181                         fl6->flowi6_proto = nexthdr;
 182                         return;
 183
 184 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 185                 case IPPROTO_MH:
 186                         if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) {
 187                                 struct ip6_mh *mh;
 188                                 mh = (struct ip6_mh *)exthdr;
 189
 190                                 fl6->fl6_mh_type = mh->ip6mh_type;
 191                         }
 192                         fl6->flowi6_proto = nexthdr;
 193                         return;
 194 #endif
 195
 196                 /* XXX Why are there these headers? */
 197                 case IPPROTO_AH:
 198                 case IPPROTO_ESP:
 199                 case IPPROTO_COMP:
 200                 default:
 201                         fl6->fl6_ipsec_spi = 0;
 202                         fl6->flowi6_proto = nexthdr;
 203                         return;
 204                 }
 205         }
 206 }
 207
 208 static inline int xfrm6_garbage_collect(struct dst_ops *ops)
 209 {
 210         struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops);
 211
 212         xfrm6_policy_afinfo.garbage_collect(net);
 213         return dst_entries_get_fast(ops) > ops->gc_thresh * 2;
 214 }
 215
 216 static void xfrm6_update_pmtu(struct dst_entry *dst, struct sock *sk,
 217                               struct sk_buff *skb, u32 mtu)
 218 {
 219         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 220         struct dst_entry *path = xdst->route;
 221
 222         path->ops->update_pmtu(path, sk, skb, mtu);
 223 }
 224
 225 static void xfrm6_redirect(struct dst_entry *dst, struct sock *sk,
 226                            struct sk_buff *skb)
 227 {
 228         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 229         struct dst_entry *path = xdst->route;
 230
 231         path->ops->redirect(path, sk, skb);
 232 }
 233
 234 static void xfrm6_dst_destroy(struct dst_entry *dst)
 235 {
 236         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 237
 238         if (likely(xdst->u.rt6.rt6i_idev))
 239                 in6_dev_put(xdst->u.rt6.rt6i_idev);
 240         dst_destroy_metrics_generic(dst);
 241         if (rt6_has_peer(&xdst->u.rt6)) {
 242                 struct inet_peer *peer = rt6_peer_ptr(&xdst->u.rt6);
 243                 inet_putpeer(peer);
 244         }
 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 = {
 276         .family =               AF_INET6,
 277         .protocol =             cpu_to_be16(ETH_P_IPV6),
 278         .gc =                   xfrm6_garbage_collect,
 279         .update_pmtu =          xfrm6_update_pmtu,
 280         .redirect =             xfrm6_redirect,
 281         .cow_metrics =          dst_cow_metrics_generic,
 282         .destroy =              xfrm6_dst_destroy,
 283         .ifdown =               xfrm6_dst_ifdown,
 284         .local_out =            __ip6_local_out,
 285         .gc_thresh =            1024,
 286 };
 287
 288 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = {
 289         .family =               AF_INET6,
 290         .dst_ops =              &xfrm6_dst_ops,
 291         .dst_lookup =           xfrm6_dst_lookup,
 292         .get_saddr =            xfrm6_get_saddr,
 293         .decode_session =       _decode_session6,
 294         .get_tos =              xfrm6_get_tos,
 295         .init_dst =             xfrm6_init_dst,
 296         .init_path =            xfrm6_init_path,
 297         .fill_dst =             xfrm6_fill_dst,
 298         .blackhole_route =      ip6_blackhole_route,
 299 };
 300
 301 static int __init xfrm6_policy_init(void)
 302 {
 303         return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo);
 304 }
 305
 306 static void xfrm6_policy_fini(void)
 307 {
 308         xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo);
 309 }
 310
 311 #ifdef CONFIG_SYSCTL
 312 static struct ctl_table xfrm6_policy_table[] = {
 313         {
 314                 .procname       = "xfrm6_gc_thresh",
 315                 .data           = &init_net.xfrm.xfrm6_dst_ops.gc_thresh,
 316                 .maxlen         = sizeof(int),
 317                 .mode           = 0644,
 318                 .proc_handler   = proc_dointvec,
 319         },
 320         { }
 321 };
 322
 323 static int __net_init xfrm6_net_init(struct net *net)
 324 {
 325         struct ctl_table *table;
 326         struct ctl_table_header *hdr;
 327
 328         table = xfrm6_policy_table;
 329         if (!net_eq(net, &init_net)) {
 330                 table = kmemdup(table, sizeof(xfrm6_policy_table), GFP_KERNEL);
 331                 if (!table)
 332                         goto err_alloc;
 333
 334                 table[0].data = &net->xfrm.xfrm6_dst_ops.gc_thresh;
 335         }
 336
 337         hdr = register_net_sysctl(net, "net/ipv6", table);
 338         if (!hdr)
 339                 goto err_reg;
 340
 341         net->ipv6.sysctl.xfrm6_hdr = hdr;
 342         return 0;
 343
 344 err_reg:
 345         if (!net_eq(net, &init_net))
 346                 kfree(table);
 347 err_alloc:
 348         return -ENOMEM;
 349 }
 350
 351 static void __net_exit xfrm6_net_exit(struct net *net)
 352 {
 353         struct ctl_table *table;
 354
 355         if (net->ipv6.sysctl.xfrm6_hdr == NULL)
 356                 return;
 357
 358         table = net->ipv6.sysctl.xfrm6_hdr->ctl_table_arg;
 359         unregister_net_sysctl_table(net->ipv6.sysctl.xfrm6_hdr);
 360         if (!net_eq(net, &init_net))
 361                 kfree(table);
 362 }
 363
 364 static struct pernet_operations xfrm6_net_ops = {
 365         .init   = xfrm6_net_init,
 366         .exit   = xfrm6_net_exit,
 367 };
 368 #endif
 369
 370 int __init xfrm6_init(void)
 371 {
 372         int ret;
 373
 374         dst_entries_init(&xfrm6_dst_ops);
 375
 376         ret = xfrm6_policy_init();
 377         if (ret) {
 378                 dst_entries_destroy(&xfrm6_dst_ops);
 379                 goto out;
 380         }
 381         ret = xfrm6_state_init();
 382         if (ret)
 383                 goto out_policy;
 384
 385 #ifdef CONFIG_SYSCTL
 386         register_pernet_subsys(&xfrm6_net_ops);
 387 #endif
 388 out:
 389         return ret;
 390 out_policy:
 391         xfrm6_policy_fini();
 392         goto out;
 393 }
 394
 395 void xfrm6_fini(void)
 396 {
 397 #ifdef CONFIG_SYSCTL
 398         unregister_pernet_subsys(&xfrm6_net_ops);
 399 #endif
 400         xfrm6_policy_fini();
 401         xfrm6_state_fini();
 402         dst_entries_destroy(&xfrm6_dst_ops);
 403 }