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 xfrm_policy_afinfo xfrm4_policy_afinfo;
  21
  22 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
  23                                             int tos, int oif,
  24                                             const xfrm_address_t *saddr,
  25                                             const xfrm_address_t *daddr)
  26 {
  27         struct rtable *rt;
  28
  29         memset(fl4, 0, sizeof(*fl4));
  30         fl4->daddr = daddr->a4;
  31         fl4->flowi4_tos = tos;
  32         fl4->flowi4_oif = oif;
  33         if (saddr)
  34                 fl4->saddr = saddr->a4;
  35
  36         fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF;
  37
  38         rt = __ip_route_output_key(net, fl4);
  39         if (!IS_ERR(rt))
  40                 return &rt->dst;
  41
  42         return ERR_CAST(rt);
  43 }
  44
  45 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif,
  46                                           const xfrm_address_t *saddr,
  47                                           const xfrm_address_t *daddr)
  48 {
  49         struct flowi4 fl4;
  50
  51         return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr);
  52 }
  53
  54 static int xfrm4_get_saddr(struct net *net, int oif,
  55                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  56 {
  57         struct dst_entry *dst;
  58         struct flowi4 fl4;
  59
  60         dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr);
  61         if (IS_ERR(dst))
  62                 return -EHOSTUNREACH;
  63
  64         saddr->a4 = fl4.saddr;
  65         dst_release(dst);
  66         return 0;
  67 }
  68
  69 static int xfrm4_get_tos(const struct flowi *fl)
  70 {
  71         return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
  72 }
  73
  74 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  75                            int nfheader_len)
  76 {
  77         return 0;
  78 }
  79
  80 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  81                           const struct flowi *fl)
  82 {
  83         struct rtable *rt = (struct rtable *)xdst->route;
  84         const struct flowi4 *fl4 = &fl->u.ip4;
  85
  86         xdst->u.rt.rt_iif = fl4->flowi4_iif;
  87
  88         xdst->u.dst.dev = dev;
  89         dev_hold(dev);
  90
  91         /* Sheit... I remember I did this right. Apparently,
  92          * it was magically lost, so this code needs audit */
  93         xdst->u.rt.rt_is_input = rt->rt_is_input;
  94         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
  95                                               RTCF_LOCAL);
  96         xdst->u.rt.rt_type = rt->rt_type;
  97         xdst->u.rt.rt_gateway = rt->rt_gateway;
  98         xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
  99         xdst->u.rt.rt_pmtu = rt->rt_pmtu;
 100         xdst->u.rt.rt_mtu_locked = rt->rt_mtu_locked;
 101         xdst->u.rt.rt_table_id = rt->rt_table_id;
 102         INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
 103
 104         return 0;
 105 }
 106
 107 static void
 108 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 109 {
 110         const struct iphdr *iph = ip_hdr(skb);
 111         int ihl = iph->ihl;
 112         u8 *xprth = skb_network_header(skb) + ihl * 4;
 113         struct flowi4 *fl4 = &fl->u.ip4;
 114         int oif = 0;
 115
 116         if (skb_dst(skb))
 117                 oif = l3mdev_fib_oif(skb_dst(skb)->dev);
 118
 119         memset(fl4, 0, sizeof(struct flowi4));
 120         fl4->flowi4_mark = skb->mark;
 121         fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 122
 123         fl4->flowi4_proto = iph->protocol;
 124         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 125         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 126         fl4->flowi4_tos = iph->tos;
 127
 128         if (!ip_is_fragment(iph)) {
 129                 switch (iph->protocol) {
 130                 case IPPROTO_UDP:
 131                 case IPPROTO_UDPLITE:
 132                 case IPPROTO_TCP:
 133                 case IPPROTO_SCTP:
 134                 case IPPROTO_DCCP:
 135                         if (xprth + 4 < skb->data ||
 136                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 137                                 __be16 *ports;
 138
 139                                 xprth = skb_network_header(skb) + ihl * 4;
 140                                 ports = (__be16 *)xprth;
 141
 142                                 fl4->fl4_sport = ports[!!reverse];
 143                                 fl4->fl4_dport = ports[!reverse];
 144                         }
 145                         break;
 146
 147                 case IPPROTO_ICMP:
 148                         if (xprth + 2 < skb->data ||
 149                             pskb_may_pull(skb, xprth + 2 - skb->data)) {
 150                                 u8 *icmp;
 151
 152                                 xprth = skb_network_header(skb) + ihl * 4;
 153                                 icmp = xprth;
 154
 155                                 fl4->fl4_icmp_type = icmp[0];
 156                                 fl4->fl4_icmp_code = icmp[1];
 157                         }
 158                         break;
 159
 160                 case IPPROTO_ESP:
 161                         if (xprth + 4 < skb->data ||
 162                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 163                                 __be32 *ehdr;
 164
 165                                 xprth = skb_network_header(skb) + ihl * 4;
 166                                 ehdr = (__be32 *)xprth;
 167
 168                                 fl4->fl4_ipsec_spi = ehdr[0];
 169                         }
 170                         break;
 171
 172                 case IPPROTO_AH:
 173                         if (xprth + 8 < skb->data ||
 174                             pskb_may_pull(skb, xprth + 8 - skb->data)) {
 175                                 __be32 *ah_hdr;
 176
 177                                 xprth = skb_network_header(skb) + ihl * 4;
 178                                 ah_hdr = (__be32 *)xprth;
 179
 180                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 181                         }
 182                         break;
 183
 184                 case IPPROTO_COMP:
 185                         if (xprth + 4 < skb->data ||
 186                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 187                                 __be16 *ipcomp_hdr;
 188
 189                                 xprth = skb_network_header(skb) + ihl * 4;
 190                                 ipcomp_hdr = (__be16 *)xprth;
 191
 192                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 193                         }
 194                         break;
 195
 196                 case IPPROTO_GRE:
 197                         if (xprth + 12 < skb->data ||
 198                             pskb_may_pull(skb, xprth + 12 - skb->data)) {
 199                                 __be16 *greflags;
 200                                 __be32 *gre_hdr;
 201
 202                                 xprth = skb_network_header(skb) + ihl * 4;
 203                                 greflags = (__be16 *)xprth;
 204                                 gre_hdr = (__be32 *)xprth;
 205
 206                                 if (greflags[0] & GRE_KEY) {
 207                                         if (greflags[0] & GRE_CSUM)
 208                                                 gre_hdr++;
 209                                         fl4->fl4_gre_key = gre_hdr[1];
 210                                 }
 211                         }
 212                         break;
 213
 214                 default:
 215                         fl4->fl4_ipsec_spi = 0;
 216                         break;
 217                 }
 218         }
 219 }
 220
 221 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 222 {
 223         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 224
 225         xfrm4_policy_afinfo.garbage_collect(net);
 226         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 227 }
 228
 229 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 230                               struct sk_buff *skb, u32 mtu)
 231 {
 232         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 233         struct dst_entry *path = xdst->route;
 234
 235         path->ops->update_pmtu(path, sk, skb, mtu);
 236 }
 237
 238 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 239                            struct sk_buff *skb)
 240 {
 241         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 242         struct dst_entry *path = xdst->route;
 243
 244         path->ops->redirect(path, sk, skb);
 245 }
 246
 247 static void xfrm4_dst_destroy(struct dst_entry *dst)
 248 {
 249         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 250
 251         dst_destroy_metrics_generic(dst);
 252
 253         xfrm_dst_destroy(xdst);
 254 }
 255
 256 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 257                              int unregister)
 258 {
 259         if (!unregister)
 260                 return;
 261
 262         xfrm_dst_ifdown(dst, dev);
 263 }
 264
 265 static struct dst_ops xfrm4_dst_ops_template = {
 266         .family =               AF_INET,
 267         .gc =                   xfrm4_garbage_collect,
 268         .update_pmtu =          xfrm4_update_pmtu,
 269         .redirect =             xfrm4_redirect,
 270         .cow_metrics =          dst_cow_metrics_generic,
 271         .destroy =              xfrm4_dst_destroy,
 272         .ifdown =               xfrm4_dst_ifdown,
 273         .local_out =            __ip_local_out,
 274         .gc_thresh =            INT_MAX,
 275 };
 276
 277 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 278         .family =               AF_INET,
 279         .dst_ops =              &xfrm4_dst_ops_template,
 280         .dst_lookup =           xfrm4_dst_lookup,
 281         .get_saddr =            xfrm4_get_saddr,
 282         .decode_session =       _decode_session4,
 283         .get_tos =              xfrm4_get_tos,
 284         .init_path =            xfrm4_init_path,
 285         .fill_dst =             xfrm4_fill_dst,
 286         .blackhole_route =      ipv4_blackhole_route,
 287 };
 288
 289 #ifdef CONFIG_SYSCTL
 290 static struct ctl_table xfrm4_policy_table[] = {
 291         {
 292                 .procname       = "xfrm4_gc_thresh",
 293                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 294                 .maxlen         = sizeof(int),
 295                 .mode           = 0644,
 296                 .proc_handler   = proc_dointvec,
 297         },
 298         { }
 299 };
 300
 301 static int __net_init xfrm4_net_sysctl_init(struct net *net)
 302 {
 303         struct ctl_table *table;
 304         struct ctl_table_header *hdr;
 305
 306         table = xfrm4_policy_table;
 307         if (!net_eq(net, &init_net)) {
 308                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 309                 if (!table)
 310                         goto err_alloc;
 311
 312                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 313         }
 314
 315         hdr = register_net_sysctl(net, "net/ipv4", table);
 316         if (!hdr)
 317                 goto err_reg;
 318
 319         net->ipv4.xfrm4_hdr = hdr;
 320         return 0;
 321
 322 err_reg:
 323         if (!net_eq(net, &init_net))
 324                 kfree(table);
 325 err_alloc:
 326         return -ENOMEM;
 327 }
 328
 329 static void __net_exit xfrm4_net_sysctl_exit(struct net *net)
 330 {
 331         struct ctl_table *table;
 332
 333         if (!net->ipv4.xfrm4_hdr)
 334                 return;
 335
 336         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 337         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 338         if (!net_eq(net, &init_net))
 339                 kfree(table);
 340 }
 341 #else /* CONFIG_SYSCTL */
 342 static int inline xfrm4_net_sysctl_init(struct net *net)
 343 {
 344         return 0;
 345 }
 346
 347 static void inline xfrm4_net_sysctl_exit(struct net *net)
 348 {
 349 }
 350 #endif
 351
 352 static int __net_init xfrm4_net_init(struct net *net)
 353 {
 354         int ret;
 355
 356         memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
 357                sizeof(xfrm4_dst_ops_template));
 358         ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
 359         if (ret)
 360                 return ret;
 361
 362         ret = xfrm4_net_sysctl_init(net);
 363         if (ret)
 364                 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 365
 366         return ret;
 367 }
 368
 369 static void __net_exit xfrm4_net_exit(struct net *net)
 370 {
 371         xfrm4_net_sysctl_exit(net);
 372         dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 373 }
 374
 375 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 376         .init   = xfrm4_net_init,
 377         .exit   = xfrm4_net_exit,
 378 };
 379
 380 static void __init xfrm4_policy_init(void)
 381 {
 382         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 383 }
 384
 385 void __init xfrm4_init(void)
 386 {
 387         xfrm4_state_init();
 388         xfrm4_policy_init();
 389         xfrm4_protocol_init();
 390         register_pernet_subsys(&xfrm4_net_ops);
 391 }
 392