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         rt_add_uncached_list(&xdst->u.rt);
 101
 102         return 0;
 103 }
 104
 105 static void
 106 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 107 {
 108         const struct iphdr *iph = ip_hdr(skb);
 109         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 110         struct flowi4 *fl4 = &fl->u.ip4;
 111         int oif = 0;
 112
 113         if (skb_dst(skb))
 114                 oif = skb_dst(skb)->dev->ifindex;
 115
 116         memset(fl4, 0, sizeof(struct flowi4));
 117         fl4->flowi4_mark = skb->mark;
 118         fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 119
 120         if (!ip_is_fragment(iph)) {
 121                 switch (iph->protocol) {
 122                 case IPPROTO_UDP:
 123                 case IPPROTO_UDPLITE:
 124                 case IPPROTO_TCP:
 125                 case IPPROTO_SCTP:
 126                 case IPPROTO_DCCP:
 127                         if (xprth + 4 < skb->data ||
 128                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 129                                 __be16 *ports;
 130
 131                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 132                                 ports = (__be16 *)xprth;
 133
 134                                 fl4->fl4_sport = ports[!!reverse];
 135                                 fl4->fl4_dport = ports[!reverse];
 136                         }
 137                         break;
 138
 139                 case IPPROTO_ICMP:
 140                         if (xprth + 2 < skb->data ||
 141                             pskb_may_pull(skb, xprth + 2 - skb->data)) {
 142                                 u8 *icmp;
 143
 144                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 145                                 icmp = xprth;
 146
 147                                 fl4->fl4_icmp_type = icmp[0];
 148                                 fl4->fl4_icmp_code = icmp[1];
 149                         }
 150                         break;
 151
 152                 case IPPROTO_ESP:
 153                         if (xprth + 4 < skb->data ||
 154                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 155                                 __be32 *ehdr;
 156
 157                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 158                                 ehdr = (__be32 *)xprth;
 159
 160                                 fl4->fl4_ipsec_spi = ehdr[0];
 161                         }
 162                         break;
 163
 164                 case IPPROTO_AH:
 165                         if (xprth + 8 < skb->data ||
 166                             pskb_may_pull(skb, xprth + 8 - skb->data)) {
 167                                 __be32 *ah_hdr;
 168
 169                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 170                                 ah_hdr = (__be32 *)xprth;
 171
 172                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 173                         }
 174                         break;
 175
 176                 case IPPROTO_COMP:
 177                         if (xprth + 4 < skb->data ||
 178                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 179                                 __be16 *ipcomp_hdr;
 180
 181                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 182                                 ipcomp_hdr = (__be16 *)xprth;
 183
 184                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 185                         }
 186                         break;
 187
 188                 case IPPROTO_GRE:
 189                         if (xprth + 12 < skb->data ||
 190                             pskb_may_pull(skb, xprth + 12 - skb->data)) {
 191                                 __be16 *greflags;
 192                                 __be32 *gre_hdr;
 193
 194                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 195                                 greflags = (__be16 *)xprth;
 196                                 gre_hdr = (__be32 *)xprth;
 197
 198                                 if (greflags[0] & GRE_KEY) {
 199                                         if (greflags[0] & GRE_CSUM)
 200                                                 gre_hdr++;
 201                                         fl4->fl4_gre_key = gre_hdr[1];
 202                                 }
 203                         }
 204                         break;
 205
 206                 default:
 207                         fl4->fl4_ipsec_spi = 0;
 208                         break;
 209                 }
 210         }
 211         fl4->flowi4_proto = iph->protocol;
 212         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 213         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 214         fl4->flowi4_tos = iph->tos;
 215 }
 216
 217 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 218 {
 219         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 220
 221         xfrm_garbage_collect_deferred(net);
 222         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 223 }
 224
 225 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 226                               struct sk_buff *skb, u32 mtu)
 227 {
 228         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 229         struct dst_entry *path = xdst->route;
 230
 231         path->ops->update_pmtu(path, sk, skb, mtu);
 232 }
 233
 234 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 235                            struct sk_buff *skb)
 236 {
 237         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 238         struct dst_entry *path = xdst->route;
 239
 240         path->ops->redirect(path, sk, skb);
 241 }
 242
 243 static void xfrm4_dst_destroy(struct dst_entry *dst)
 244 {
 245         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 246
 247         dst_destroy_metrics_generic(dst);
 248         if (xdst->u.rt.rt_uncached_list)
 249                 rt_del_uncached_list(&xdst->u.rt);
 250         xfrm_dst_destroy(xdst);
 251 }
 252
 253 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 254                              int unregister)
 255 {
 256         if (!unregister)
 257                 return;
 258
 259         xfrm_dst_ifdown(dst, dev);
 260 }
 261
 262 static struct dst_ops xfrm4_dst_ops_template = {
 263         .family =               AF_INET,
 264         .gc =                   xfrm4_garbage_collect,
 265         .update_pmtu =          xfrm4_update_pmtu,
 266         .redirect =             xfrm4_redirect,
 267         .cow_metrics =          dst_cow_metrics_generic,
 268         .destroy =              xfrm4_dst_destroy,
 269         .ifdown =               xfrm4_dst_ifdown,
 270         .local_out =            __ip_local_out,
 271         .gc_thresh =            INT_MAX,
 272 };
 273
 274 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 275         .dst_ops =              &xfrm4_dst_ops_template,
 276         .dst_lookup =           xfrm4_dst_lookup,
 277         .get_saddr =            xfrm4_get_saddr,
 278         .decode_session =       _decode_session4,
 279         .get_tos =              xfrm4_get_tos,
 280         .init_path =            xfrm4_init_path,
 281         .fill_dst =             xfrm4_fill_dst,
 282         .blackhole_route =      ipv4_blackhole_route,
 283 };
 284
 285 #ifdef CONFIG_SYSCTL
 286 static struct ctl_table xfrm4_policy_table[] = {
 287         {
 288                 .procname       = "xfrm4_gc_thresh",
 289                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 290                 .maxlen         = sizeof(int),
 291                 .mode           = 0644,
 292                 .proc_handler   = proc_dointvec,
 293         },
 294         { }
 295 };
 296
 297 static __net_init int xfrm4_net_sysctl_init(struct net *net)
 298 {
 299         struct ctl_table *table;
 300         struct ctl_table_header *hdr;
 301
 302         table = xfrm4_policy_table;
 303         if (!net_eq(net, &init_net)) {
 304                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 305                 if (!table)
 306                         goto err_alloc;
 307
 308                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 309         }
 310
 311         hdr = register_net_sysctl(net, "net/ipv4", table);
 312         if (!hdr)
 313                 goto err_reg;
 314
 315         net->ipv4.xfrm4_hdr = hdr;
 316         return 0;
 317
 318 err_reg:
 319         if (!net_eq(net, &init_net))
 320                 kfree(table);
 321 err_alloc:
 322         return -ENOMEM;
 323 }
 324
 325 static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
 326 {
 327         struct ctl_table *table;
 328
 329         if (!net->ipv4.xfrm4_hdr)
 330                 return;
 331
 332         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 333         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 334         if (!net_eq(net, &init_net))
 335                 kfree(table);
 336 }
 337 #else /* CONFIG_SYSCTL */
 338 static inline int xfrm4_net_sysctl_init(struct net *net)
 339 {
 340         return 0;
 341 }
 342
 343 static inline void xfrm4_net_sysctl_exit(struct net *net)
 344 {
 345 }
 346 #endif
 347
 348 static int __net_init xfrm4_net_init(struct net *net)
 349 {
 350         int ret;
 351
 352         memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
 353                sizeof(xfrm4_dst_ops_template));
 354         ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
 355         if (ret)
 356                 return ret;
 357
 358         ret = xfrm4_net_sysctl_init(net);
 359         if (ret)
 360                 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 361
 362         return ret;
 363 }
 364
 365 static void __net_exit xfrm4_net_exit(struct net *net)
 366 {
 367         xfrm4_net_sysctl_exit(net);
 368         dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 369 }
 370
 371 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 372         .init   = xfrm4_net_init,
 373         .exit   = xfrm4_net_exit,
 374 };
 375
 376 static void __init xfrm4_policy_init(void)
 377 {
 378         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
 379 }
 380
 381 void __init xfrm4_init(void)
 382 {
 383         xfrm4_state_init();
 384         xfrm4_policy_init();
 385         xfrm4_protocol_init();
 386         register_pernet_subsys(&xfrm4_net_ops);
 387 }
 388