net/ipv4/xfrm4_policy.c

   1 /*
   2  * xfrm4_policy.c
   3  *
   4  * Changes:
   5  *      Kazunori MIYAZAWA @USAGI
   6  *      YOSHIFUJI Hideaki @USAGI
   7  *              Split up af-specific portion
   8  *
   9  */
  10
  11 #include <linux/err.h>
  12 #include <linux/kernel.h>
  13 #include <linux/inetdevice.h>
  14 #include <linux/if_tunnel.h>
  15 #include <net/dst.h>
  16 #include <net/xfrm.h>
  17 #include <net/ip.h>
  18 #include <net/l3mdev.h>
  19
  20 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
  21                                             int tos, int oif,
  22                                             const xfrm_address_t *saddr,
  23                                             const xfrm_address_t *daddr)
  24 {
  25         struct rtable *rt;
  26
  27         memset(fl4, 0, sizeof(*fl4));
  28         fl4->daddr = daddr->a4;
  29         fl4->flowi4_tos = tos;
  30         fl4->flowi4_oif = l3mdev_master_ifindex_by_index(net, oif);
  31         if (saddr)
  32                 fl4->saddr = saddr->a4;
  33
  34         fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF;
  35
  36         rt = __ip_route_output_key(net, fl4);
  37         if (!IS_ERR(rt))
  38                 return &rt->dst;
  39
  40         return ERR_CAST(rt);
  41 }
  42
  43 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif,
  44                                           const xfrm_address_t *saddr,
  45                                           const xfrm_address_t *daddr)
  46 {
  47         struct flowi4 fl4;
  48
  49         return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr);
  50 }
  51
  52 static int xfrm4_get_saddr(struct net *net, int oif,
  53                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  54 {
  55         struct dst_entry *dst;
  56         struct flowi4 fl4;
  57
  58         dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr);
  59         if (IS_ERR(dst))
  60                 return -EHOSTUNREACH;
  61
  62         saddr->a4 = fl4.saddr;
  63         dst_release(dst);
  64         return 0;
  65 }
  66
  67 static int xfrm4_get_tos(const struct flowi *fl)
  68 {
  69         return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
  70 }
  71
  72 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  73                            int nfheader_len)
  74 {
  75         return 0;
  76 }
  77
  78 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  79                           const struct flowi *fl)
  80 {
  81         struct rtable *rt = (struct rtable *)xdst->route;
  82         const struct flowi4 *fl4 = &fl->u.ip4;
  83
  84         xdst->u.rt.rt_iif = fl4->flowi4_iif;
  85
  86         xdst->u.dst.dev = dev;
  87         dev_hold(dev);
  88
  89         /* Sheit... I remember I did this right. Apparently,
  90          * it was magically lost, so this code needs audit */
  91         xdst->u.rt.rt_is_input = rt->rt_is_input;
  92         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
  93                                               RTCF_LOCAL);
  94         xdst->u.rt.rt_type = rt->rt_type;
  95         xdst->u.rt.rt_gateway = rt->rt_gateway;
  96         xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
  97         xdst->u.rt.rt_pmtu = rt->rt_pmtu;
  98         xdst->u.rt.rt_table_id = rt->rt_table_id;
  99         INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
 100
 101         return 0;
 102 }
 103
 104 static void
 105 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 106 {
 107         const struct iphdr *iph = ip_hdr(skb);
 108         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 109         struct flowi4 *fl4 = &fl->u.ip4;
 110         int oif = 0;
 111
 112         if (skb_dst(skb))
 113                 oif = skb_dst(skb)->dev->ifindex;
 114
 115         memset(fl4, 0, sizeof(struct flowi4));
 116         fl4->flowi4_mark = skb->mark;
 117         fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 118
 119         if (!ip_is_fragment(iph)) {
 120                 switch (iph->protocol) {
 121                 case IPPROTO_UDP:
 122                 case IPPROTO_UDPLITE:
 123                 case IPPROTO_TCP:
 124                 case IPPROTO_SCTP:
 125                 case IPPROTO_DCCP:
 126                         if (xprth + 4 < skb->data ||
 127                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 128                                 __be16 *ports;
 129
 130                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 131                                 ports = (__be16 *)xprth;
 132
 133                                 fl4->fl4_sport = ports[!!reverse];
 134                                 fl4->fl4_dport = ports[!reverse];
 135                         }
 136                         break;
 137
 138                 case IPPROTO_ICMP:
 139                         if (xprth + 2 < skb->data ||
 140                             pskb_may_pull(skb, xprth + 2 - skb->data)) {
 141                                 u8 *icmp;
 142
 143                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 144                                 icmp = xprth;
 145
 146                                 fl4->fl4_icmp_type = icmp[0];
 147                                 fl4->fl4_icmp_code = icmp[1];
 148                         }
 149                         break;
 150
 151                 case IPPROTO_ESP:
 152                         if (xprth + 4 < skb->data ||
 153                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 154                                 __be32 *ehdr;
 155
 156                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 157                                 ehdr = (__be32 *)xprth;
 158
 159                                 fl4->fl4_ipsec_spi = ehdr[0];
 160                         }
 161                         break;
 162
 163                 case IPPROTO_AH:
 164                         if (xprth + 8 < skb->data ||
 165                             pskb_may_pull(skb, xprth + 8 - skb->data)) {
 166                                 __be32 *ah_hdr;
 167
 168                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 169                                 ah_hdr = (__be32 *)xprth;
 170
 171                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 172                         }
 173                         break;
 174
 175                 case IPPROTO_COMP:
 176                         if (xprth + 4 < skb->data ||
 177                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 178                                 __be16 *ipcomp_hdr;
 179
 180                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 181                                 ipcomp_hdr = (__be16 *)xprth;
 182
 183                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 184                         }
 185                         break;
 186
 187                 case IPPROTO_GRE:
 188                         if (xprth + 12 < skb->data ||
 189                             pskb_may_pull(skb, xprth + 12 - skb->data)) {
 190                                 __be16 *greflags;
 191                                 __be32 *gre_hdr;
 192
 193                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 194                                 greflags = (__be16 *)xprth;
 195                                 gre_hdr = (__be32 *)xprth;
 196
 197                                 if (greflags[0] & GRE_KEY) {
 198                                         if (greflags[0] & GRE_CSUM)
 199                                                 gre_hdr++;
 200                                         fl4->fl4_gre_key = gre_hdr[1];
 201                                 }
 202                         }
 203                         break;
 204
 205                 default:
 206                         fl4->fl4_ipsec_spi = 0;
 207                         break;
 208                 }
 209         }
 210         fl4->flowi4_proto = iph->protocol;
 211         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 212         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 213         fl4->flowi4_tos = iph->tos;
 214 }
 215
 216 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 217 {
 218         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 219
 220         xfrm_garbage_collect_deferred(net);
 221         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 222 }
 223
 224 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 225                               struct sk_buff *skb, u32 mtu)
 226 {
 227         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 228         struct dst_entry *path = xdst->route;
 229
 230         path->ops->update_pmtu(path, sk, skb, mtu);
 231 }
 232
 233 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 234                            struct sk_buff *skb)
 235 {
 236         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 237         struct dst_entry *path = xdst->route;
 238
 239         path->ops->redirect(path, sk, skb);
 240 }
 241
 242 static void xfrm4_dst_destroy(struct dst_entry *dst)
 243 {
 244         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 245
 246         dst_destroy_metrics_generic(dst);
 247
 248         xfrm_dst_destroy(xdst);
 249 }
 250
 251 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 252                              int unregister)
 253 {
 254         if (!unregister)
 255                 return;
 256
 257         xfrm_dst_ifdown(dst, dev);
 258 }
 259
 260 static struct dst_ops xfrm4_dst_ops_template = {
 261         .family =               AF_INET,
 262         .gc =                   xfrm4_garbage_collect,
 263         .update_pmtu =          xfrm4_update_pmtu,
 264         .redirect =             xfrm4_redirect,
 265         .cow_metrics =          dst_cow_metrics_generic,
 266         .destroy =              xfrm4_dst_destroy,
 267         .ifdown =               xfrm4_dst_ifdown,
 268         .local_out =            __ip_local_out,
 269         .gc_thresh =            INT_MAX,
 270 };
 271
 272 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 273         .dst_ops =              &xfrm4_dst_ops_template,
 274         .dst_lookup =           xfrm4_dst_lookup,
 275         .get_saddr =            xfrm4_get_saddr,
 276         .decode_session =       _decode_session4,
 277         .get_tos =              xfrm4_get_tos,
 278         .init_path =            xfrm4_init_path,
 279         .fill_dst =             xfrm4_fill_dst,
 280         .blackhole_route =      ipv4_blackhole_route,
 281 };
 282
 283 #ifdef CONFIG_SYSCTL
 284 static struct ctl_table xfrm4_policy_table[] = {
 285         {
 286                 .procname       = "xfrm4_gc_thresh",
 287                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 288                 .maxlen         = sizeof(int),
 289                 .mode           = 0644,
 290                 .proc_handler   = proc_dointvec,
 291         },
 292         { }
 293 };
 294
 295 static __net_init int xfrm4_net_sysctl_init(struct net *net)
 296 {
 297         struct ctl_table *table;
 298         struct ctl_table_header *hdr;
 299
 300         table = xfrm4_policy_table;
 301         if (!net_eq(net, &init_net)) {
 302                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 303                 if (!table)
 304                         goto err_alloc;
 305
 306                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 307         }
 308
 309         hdr = register_net_sysctl(net, "net/ipv4", table);
 310         if (!hdr)
 311                 goto err_reg;
 312
 313         net->ipv4.xfrm4_hdr = hdr;
 314         return 0;
 315
 316 err_reg:
 317         if (!net_eq(net, &init_net))
 318                 kfree(table);
 319 err_alloc:
 320         return -ENOMEM;
 321 }
 322
 323 static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
 324 {
 325         struct ctl_table *table;
 326
 327         if (!net->ipv4.xfrm4_hdr)
 328                 return;
 329
 330         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 331         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 332         if (!net_eq(net, &init_net))
 333                 kfree(table);
 334 }
 335 #else /* CONFIG_SYSCTL */
 336 static inline int xfrm4_net_sysctl_init(struct net *net)
 337 {
 338         return 0;
 339 }
 340
 341 static inline void xfrm4_net_sysctl_exit(struct net *net)
 342 {
 343 }
 344 #endif
 345
 346 static int __net_init xfrm4_net_init(struct net *net)
 347 {
 348         int ret;
 349
 350         memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
 351                sizeof(xfrm4_dst_ops_template));
 352         ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
 353         if (ret)
 354                 return ret;
 355
 356         ret = xfrm4_net_sysctl_init(net);
 357         if (ret)
 358                 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 359
 360         return ret;
 361 }
 362
 363 static void __net_exit xfrm4_net_exit(struct net *net)
 364 {
 365         xfrm4_net_sysctl_exit(net);
 366         dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 367 }
 368
 369 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 370         .init   = xfrm4_net_init,
 371         .exit   = xfrm4_net_exit,
 372 };
 373
 374 static void __init xfrm4_policy_init(void)
 375 {
 376         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
 377 }
 378
 379 void __init xfrm4_init(void)
 380 {
 381         xfrm4_state_init();
 382         xfrm4_policy_init();
 383         xfrm4_protocol_init();
 384         register_pernet_subsys(&xfrm4_net_ops);
 385 }
 386