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                                             u32 mark)
  25 {
  26         struct rtable *rt;
  27
  28         memset(fl4, 0, sizeof(*fl4));
  29         fl4->daddr = daddr->a4;
  30         fl4->flowi4_tos = tos;
  31         fl4->flowi4_oif = l3mdev_master_ifindex_by_index(net, oif);
  32         fl4->flowi4_mark = mark;
  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                                           u32 mark)
  49 {
  50         struct flowi4 fl4;
  51
  52         return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr, mark);
  53 }
  54
  55 static int xfrm4_get_saddr(struct net *net, int oif,
  56                            xfrm_address_t *saddr, xfrm_address_t *daddr,
  57                            u32 mark)
  58 {
  59         struct dst_entry *dst;
  60         struct flowi4 fl4;
  61
  62         dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr, mark);
  63         if (IS_ERR(dst))
  64                 return -EHOSTUNREACH;
  65
  66         saddr->a4 = fl4.saddr;
  67         dst_release(dst);
  68         return 0;
  69 }
  70
  71 static int xfrm4_get_tos(const struct flowi *fl)
  72 {
  73         return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
  74 }
  75
  76 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  77                            int nfheader_len)
  78 {
  79         return 0;
  80 }
  81
  82 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  83                           const struct flowi *fl)
  84 {
  85         struct rtable *rt = (struct rtable *)xdst->route;
  86         const struct flowi4 *fl4 = &fl->u.ip4;
  87
  88         xdst->u.rt.rt_iif = fl4->flowi4_iif;
  89
  90         xdst->u.dst.dev = dev;
  91         dev_hold(dev);
  92
  93         /* Sheit... I remember I did this right. Apparently,
  94          * it was magically lost, so this code needs audit */
  95         xdst->u.rt.rt_is_input = rt->rt_is_input;
  96         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
  97                                               RTCF_LOCAL);
  98         xdst->u.rt.rt_type = rt->rt_type;
  99         xdst->u.rt.rt_gateway = rt->rt_gateway;
 100         xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
 101         xdst->u.rt.rt_pmtu = rt->rt_pmtu;
 102         xdst->u.rt.rt_table_id = rt->rt_table_id;
 103         INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
 104
 105         return 0;
 106 }
 107
 108 static void
 109 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 110 {
 111         const struct iphdr *iph = ip_hdr(skb);
 112         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 113         struct flowi4 *fl4 = &fl->u.ip4;
 114         int oif = 0;
 115
 116         if (skb_dst(skb))
 117                 oif = skb_dst(skb)->dev->ifindex;
 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         if (!ip_is_fragment(iph)) {
 124                 switch (iph->protocol) {
 125                 case IPPROTO_UDP:
 126                 case IPPROTO_UDPLITE:
 127                 case IPPROTO_TCP:
 128                 case IPPROTO_SCTP:
 129                 case IPPROTO_DCCP:
 130                         if (xprth + 4 < skb->data ||
 131                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 132                                 __be16 *ports;
 133
 134                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 135                                 ports = (__be16 *)xprth;
 136
 137                                 fl4->fl4_sport = ports[!!reverse];
 138                                 fl4->fl4_dport = ports[!reverse];
 139                         }
 140                         break;
 141
 142                 case IPPROTO_ICMP:
 143                         if (xprth + 2 < skb->data ||
 144                             pskb_may_pull(skb, xprth + 2 - skb->data)) {
 145                                 u8 *icmp;
 146
 147                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 148                                 icmp = xprth;
 149
 150                                 fl4->fl4_icmp_type = icmp[0];
 151                                 fl4->fl4_icmp_code = icmp[1];
 152                         }
 153                         break;
 154
 155                 case IPPROTO_ESP:
 156                         if (xprth + 4 < skb->data ||
 157                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 158                                 __be32 *ehdr;
 159
 160                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 161                                 ehdr = (__be32 *)xprth;
 162
 163                                 fl4->fl4_ipsec_spi = ehdr[0];
 164                         }
 165                         break;
 166
 167                 case IPPROTO_AH:
 168                         if (xprth + 8 < skb->data ||
 169                             pskb_may_pull(skb, xprth + 8 - skb->data)) {
 170                                 __be32 *ah_hdr;
 171
 172                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 173                                 ah_hdr = (__be32 *)xprth;
 174
 175                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 176                         }
 177                         break;
 178
 179                 case IPPROTO_COMP:
 180                         if (xprth + 4 < skb->data ||
 181                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 182                                 __be16 *ipcomp_hdr;
 183
 184                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 185                                 ipcomp_hdr = (__be16 *)xprth;
 186
 187                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 188                         }
 189                         break;
 190
 191                 case IPPROTO_GRE:
 192                         if (xprth + 12 < skb->data ||
 193                             pskb_may_pull(skb, xprth + 12 - skb->data)) {
 194                                 __be16 *greflags;
 195                                 __be32 *gre_hdr;
 196
 197                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 198                                 greflags = (__be16 *)xprth;
 199                                 gre_hdr = (__be32 *)xprth;
 200
 201                                 if (greflags[0] & GRE_KEY) {
 202                                         if (greflags[0] & GRE_CSUM)
 203                                                 gre_hdr++;
 204                                         fl4->fl4_gre_key = gre_hdr[1];
 205                                 }
 206                         }
 207                         break;
 208
 209                 default:
 210                         fl4->fl4_ipsec_spi = 0;
 211                         break;
 212                 }
 213         }
 214         fl4->flowi4_proto = iph->protocol;
 215         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 216         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 217         fl4->flowi4_tos = iph->tos;
 218 }
 219
 220 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 221                               struct sk_buff *skb, u32 mtu)
 222 {
 223         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 224         struct dst_entry *path = xdst->route;
 225
 226         path->ops->update_pmtu(path, sk, skb, mtu);
 227 }
 228
 229 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 230                            struct sk_buff *skb)
 231 {
 232         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 233         struct dst_entry *path = xdst->route;
 234
 235         path->ops->redirect(path, sk, skb);
 236 }
 237
 238 static void xfrm4_dst_destroy(struct dst_entry *dst)
 239 {
 240         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 241
 242         dst_destroy_metrics_generic(dst);
 243
 244         xfrm_dst_destroy(xdst);
 245 }
 246
 247 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 248                              int unregister)
 249 {
 250         if (!unregister)
 251                 return;
 252
 253         xfrm_dst_ifdown(dst, dev);
 254 }
 255
 256 static struct dst_ops xfrm4_dst_ops_template = {
 257         .family =               AF_INET,
 258         .update_pmtu =          xfrm4_update_pmtu,
 259         .redirect =             xfrm4_redirect,
 260         .cow_metrics =          dst_cow_metrics_generic,
 261         .destroy =              xfrm4_dst_destroy,
 262         .ifdown =               xfrm4_dst_ifdown,
 263         .local_out =            __ip_local_out,
 264         .gc_thresh =            32768,
 265 };
 266
 267 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 268         .dst_ops =              &xfrm4_dst_ops_template,
 269         .dst_lookup =           xfrm4_dst_lookup,
 270         .get_saddr =            xfrm4_get_saddr,
 271         .decode_session =       _decode_session4,
 272         .get_tos =              xfrm4_get_tos,
 273         .init_path =            xfrm4_init_path,
 274         .fill_dst =             xfrm4_fill_dst,
 275         .blackhole_route =      ipv4_blackhole_route,
 276 };
 277
 278 #ifdef CONFIG_SYSCTL
 279 static struct ctl_table xfrm4_policy_table[] = {
 280         {
 281                 .procname       = "xfrm4_gc_thresh",
 282                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 283                 .maxlen         = sizeof(int),
 284                 .mode           = 0644,
 285                 .proc_handler   = proc_dointvec,
 286         },
 287         { }
 288 };
 289
 290 static __net_init int xfrm4_net_sysctl_init(struct net *net)
 291 {
 292         struct ctl_table *table;
 293         struct ctl_table_header *hdr;
 294
 295         table = xfrm4_policy_table;
 296         if (!net_eq(net, &init_net)) {
 297                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 298                 if (!table)
 299                         goto err_alloc;
 300
 301                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 302         }
 303
 304         hdr = register_net_sysctl(net, "net/ipv4", table);
 305         if (!hdr)
 306                 goto err_reg;
 307
 308         net->ipv4.xfrm4_hdr = hdr;
 309         return 0;
 310
 311 err_reg:
 312         if (!net_eq(net, &init_net))
 313                 kfree(table);
 314 err_alloc:
 315         return -ENOMEM;
 316 }
 317
 318 static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
 319 {
 320         struct ctl_table *table;
 321
 322         if (!net->ipv4.xfrm4_hdr)
 323                 return;
 324
 325         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 326         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 327         if (!net_eq(net, &init_net))
 328                 kfree(table);
 329 }
 330 #else /* CONFIG_SYSCTL */
 331 static inline int xfrm4_net_sysctl_init(struct net *net)
 332 {
 333         return 0;
 334 }
 335
 336 static inline void xfrm4_net_sysctl_exit(struct net *net)
 337 {
 338 }
 339 #endif
 340
 341 static int __net_init xfrm4_net_init(struct net *net)
 342 {
 343         int ret;
 344
 345         memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
 346                sizeof(xfrm4_dst_ops_template));
 347         ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
 348         if (ret)
 349                 return ret;
 350
 351         ret = xfrm4_net_sysctl_init(net);
 352         if (ret)
 353                 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 354
 355         return ret;
 356 }
 357
 358 static void __net_exit xfrm4_net_exit(struct net *net)
 359 {
 360         xfrm4_net_sysctl_exit(net);
 361         dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 362 }
 363
 364 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 365         .init   = xfrm4_net_init,
 366         .exit   = xfrm4_net_exit,
 367 };
 368
 369 static void __init xfrm4_policy_init(void)
 370 {
 371         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
 372 }
 373
 374 void __init xfrm4_init(void)
 375 {
 376         xfrm4_state_init();
 377         xfrm4_policy_init();
 378         xfrm4_protocol_init();
 379         register_pernet_subsys(&xfrm4_net_ops);
 380 }
 381