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         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 112         struct flowi4 *fl4 = &fl->u.ip4;
 113         int oif = 0;
 114
 115         if (skb_dst(skb))
 116                 oif = l3mdev_fib_oif(skb_dst(skb)->dev);
 117
 118         memset(fl4, 0, sizeof(struct flowi4));
 119         fl4->flowi4_mark = skb->mark;
 120         fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 121
 122         if (!ip_is_fragment(iph)) {
 123                 switch (iph->protocol) {
 124                 case IPPROTO_UDP:
 125                 case IPPROTO_UDPLITE:
 126                 case IPPROTO_TCP:
 127                 case IPPROTO_SCTP:
 128                 case IPPROTO_DCCP:
 129                         if (xprth + 4 < skb->data ||
 130                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 131                                 __be16 *ports;
 132
 133                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 134                                 ports = (__be16 *)xprth;
 135
 136                                 fl4->fl4_sport = ports[!!reverse];
 137                                 fl4->fl4_dport = ports[!reverse];
 138                         }
 139                         break;
 140
 141                 case IPPROTO_ICMP:
 142                         if (xprth + 2 < skb->data ||
 143                             pskb_may_pull(skb, xprth + 2 - skb->data)) {
 144                                 u8 *icmp;
 145
 146                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 147                                 icmp = xprth;
 148
 149                                 fl4->fl4_icmp_type = icmp[0];
 150                                 fl4->fl4_icmp_code = icmp[1];
 151                         }
 152                         break;
 153
 154                 case IPPROTO_ESP:
 155                         if (xprth + 4 < skb->data ||
 156                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 157                                 __be32 *ehdr;
 158
 159                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 160                                 ehdr = (__be32 *)xprth;
 161
 162                                 fl4->fl4_ipsec_spi = ehdr[0];
 163                         }
 164                         break;
 165
 166                 case IPPROTO_AH:
 167                         if (xprth + 8 < skb->data ||
 168                             pskb_may_pull(skb, xprth + 8 - skb->data)) {
 169                                 __be32 *ah_hdr;
 170
 171                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 172                                 ah_hdr = (__be32 *)xprth;
 173
 174                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 175                         }
 176                         break;
 177
 178                 case IPPROTO_COMP:
 179                         if (xprth + 4 < skb->data ||
 180                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 181                                 __be16 *ipcomp_hdr;
 182
 183                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 184                                 ipcomp_hdr = (__be16 *)xprth;
 185
 186                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 187                         }
 188                         break;
 189
 190                 case IPPROTO_GRE:
 191                         if (xprth + 12 < skb->data ||
 192                             pskb_may_pull(skb, xprth + 12 - skb->data)) {
 193                                 __be16 *greflags;
 194                                 __be32 *gre_hdr;
 195
 196                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 197                                 greflags = (__be16 *)xprth;
 198                                 gre_hdr = (__be32 *)xprth;
 199
 200                                 if (greflags[0] & GRE_KEY) {
 201                                         if (greflags[0] & GRE_CSUM)
 202                                                 gre_hdr++;
 203                                         fl4->fl4_gre_key = gre_hdr[1];
 204                                 }
 205                         }
 206                         break;
 207
 208                 default:
 209                         fl4->fl4_ipsec_spi = 0;
 210                         break;
 211                 }
 212         }
 213         fl4->flowi4_proto = iph->protocol;
 214         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 215         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 216         fl4->flowi4_tos = iph->tos;
 217 }
 218
 219 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 220 {
 221         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 222
 223         xfrm4_policy_afinfo.garbage_collect(net);
 224         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 225 }
 226
 227 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 228                               struct sk_buff *skb, u32 mtu)
 229 {
 230         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 231         struct dst_entry *path = xdst->route;
 232
 233         path->ops->update_pmtu(path, sk, skb, mtu);
 234 }
 235
 236 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 237                            struct sk_buff *skb)
 238 {
 239         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 240         struct dst_entry *path = xdst->route;
 241
 242         path->ops->redirect(path, sk, skb);
 243 }
 244
 245 static void xfrm4_dst_destroy(struct dst_entry *dst)
 246 {
 247         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 248
 249         dst_destroy_metrics_generic(dst);
 250
 251         xfrm_dst_destroy(xdst);
 252 }
 253
 254 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 255                              int unregister)
 256 {
 257         if (!unregister)
 258                 return;
 259
 260         xfrm_dst_ifdown(dst, dev);
 261 }
 262
 263 static struct dst_ops xfrm4_dst_ops_template = {
 264         .family =               AF_INET,
 265         .gc =                   xfrm4_garbage_collect,
 266         .update_pmtu =          xfrm4_update_pmtu,
 267         .redirect =             xfrm4_redirect,
 268         .cow_metrics =          dst_cow_metrics_generic,
 269         .destroy =              xfrm4_dst_destroy,
 270         .ifdown =               xfrm4_dst_ifdown,
 271         .local_out =            __ip_local_out,
 272         .gc_thresh =            INT_MAX,
 273 };
 274
 275 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 276         .family =               AF_INET,
 277         .dst_ops =              &xfrm4_dst_ops_template,
 278         .dst_lookup =           xfrm4_dst_lookup,
 279         .get_saddr =            xfrm4_get_saddr,
 280         .decode_session =       _decode_session4,
 281         .get_tos =              xfrm4_get_tos,
 282         .init_path =            xfrm4_init_path,
 283         .fill_dst =             xfrm4_fill_dst,
 284         .blackhole_route =      ipv4_blackhole_route,
 285 };
 286
 287 #ifdef CONFIG_SYSCTL
 288 static struct ctl_table xfrm4_policy_table[] = {
 289         {
 290                 .procname       = "xfrm4_gc_thresh",
 291                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 292                 .maxlen         = sizeof(int),
 293                 .mode           = 0644,
 294                 .proc_handler   = proc_dointvec,
 295         },
 296         { }
 297 };
 298
 299 static int __net_init xfrm4_net_sysctl_init(struct net *net)
 300 {
 301         struct ctl_table *table;
 302         struct ctl_table_header *hdr;
 303
 304         table = xfrm4_policy_table;
 305         if (!net_eq(net, &init_net)) {
 306                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 307                 if (!table)
 308                         goto err_alloc;
 309
 310                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 311         }
 312
 313         hdr = register_net_sysctl(net, "net/ipv4", table);
 314         if (!hdr)
 315                 goto err_reg;
 316
 317         net->ipv4.xfrm4_hdr = hdr;
 318         return 0;
 319
 320 err_reg:
 321         if (!net_eq(net, &init_net))
 322                 kfree(table);
 323 err_alloc:
 324         return -ENOMEM;
 325 }
 326
 327 static void __net_exit xfrm4_net_sysctl_exit(struct net *net)
 328 {
 329         struct ctl_table *table;
 330
 331         if (!net->ipv4.xfrm4_hdr)
 332                 return;
 333
 334         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 335         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 336         if (!net_eq(net, &init_net))
 337                 kfree(table);
 338 }
 339 #else /* CONFIG_SYSCTL */
 340 static int inline xfrm4_net_sysctl_init(struct net *net)
 341 {
 342         return 0;
 343 }
 344
 345 static void inline xfrm4_net_sysctl_exit(struct net *net)
 346 {
 347 }
 348 #endif
 349
 350 static int __net_init xfrm4_net_init(struct net *net)
 351 {
 352         int ret;
 353
 354         memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
 355                sizeof(xfrm4_dst_ops_template));
 356         ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
 357         if (ret)
 358                 return ret;
 359
 360         ret = xfrm4_net_sysctl_init(net);
 361         if (ret)
 362                 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 363
 364         return ret;
 365 }
 366
 367 static void __net_exit xfrm4_net_exit(struct net *net)
 368 {
 369         xfrm4_net_sysctl_exit(net);
 370         dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 371 }
 372
 373 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 374         .init   = xfrm4_net_init,
 375         .exit   = xfrm4_net_exit,
 376 };
 377
 378 static void __init xfrm4_policy_init(void)
 379 {
 380         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 381 }
 382
 383 void __init xfrm4_init(void)
 384 {
 385         xfrm4_state_init();
 386         xfrm4_policy_init();
 387         xfrm4_protocol_init();
 388         register_pernet_subsys(&xfrm4_net_ops);
 389 }
 390