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
  19 static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
  20
  21 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4,
  22                                             int tos,
  23                                             const xfrm_address_t *saddr,
  24                                             const xfrm_address_t *daddr)
  25 {
  26         struct rtable *rt;
  27
  28         memset(fl4, 0, sizeof(*fl4));
  29         fl4->daddr = daddr->a4;
  30         fl4->flowi4_tos = tos;
  31         if (saddr)
  32                 fl4->saddr = saddr->a4;
  33
  34         rt = __ip_route_output_key(net, fl4);
  35         if (!IS_ERR(rt))
  36                 return &rt->dst;
  37
  38         return ERR_CAST(rt);
  39 }
  40
  41 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos,
  42                                           const xfrm_address_t *saddr,
  43                                           const xfrm_address_t *daddr)
  44 {
  45         struct flowi4 fl4;
  46
  47         return __xfrm4_dst_lookup(net, &fl4, tos, saddr, daddr);
  48 }
  49
  50 static int xfrm4_get_saddr(struct net *net,
  51                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  52 {
  53         struct dst_entry *dst;
  54         struct flowi4 fl4;
  55
  56         dst = __xfrm4_dst_lookup(net, &fl4, 0, NULL, daddr);
  57         if (IS_ERR(dst))
  58                 return -EHOSTUNREACH;
  59
  60         saddr->a4 = fl4.saddr;
  61         dst_release(dst);
  62         return 0;
  63 }
  64
  65 static int xfrm4_get_tos(const struct flowi *fl)
  66 {
  67         return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */
  68 }
  69
  70 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  71                            int nfheader_len)
  72 {
  73         return 0;
  74 }
  75
  76 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  77                           const struct flowi *fl)
  78 {
  79         struct rtable *rt = (struct rtable *)xdst->route;
  80         const struct flowi4 *fl4 = &fl->u.ip4;
  81
  82         xdst->u.rt.rt_iif = fl4->flowi4_iif;
  83
  84         xdst->u.dst.dev = dev;
  85         dev_hold(dev);
  86
  87         /* Sheit... I remember I did this right. Apparently,
  88          * it was magically lost, so this code needs audit */
  89         xdst->u.rt.rt_is_input = rt->rt_is_input;
  90         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
  91                                               RTCF_LOCAL);
  92         xdst->u.rt.rt_type = rt->rt_type;
  93         xdst->u.rt.rt_gateway = rt->rt_gateway;
  94         xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway;
  95         xdst->u.rt.rt_pmtu = rt->rt_pmtu;
  96         INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
  97
  98         return 0;
  99 }
 100
 101 static void
 102 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 103 {
 104         const struct iphdr *iph = ip_hdr(skb);
 105         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 106         struct flowi4 *fl4 = &fl->u.ip4;
 107         int oif = 0;
 108
 109         if (skb_dst(skb))
 110                 oif = skb_dst(skb)->dev->ifindex;
 111
 112         memset(fl4, 0, sizeof(struct flowi4));
 113         fl4->flowi4_mark = skb->mark;
 114         fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 115
 116         if (!ip_is_fragment(iph)) {
 117                 switch (iph->protocol) {
 118                 case IPPROTO_UDP:
 119                 case IPPROTO_UDPLITE:
 120                 case IPPROTO_TCP:
 121                 case IPPROTO_SCTP:
 122                 case IPPROTO_DCCP:
 123                         if (xprth + 4 < skb->data ||
 124                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 125                                 __be16 *ports = (__be16 *)xprth;
 126
 127                                 fl4->fl4_sport = ports[!!reverse];
 128                                 fl4->fl4_dport = ports[!reverse];
 129                         }
 130                         break;
 131
 132                 case IPPROTO_ICMP:
 133                         if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
 134                                 u8 *icmp = xprth;
 135
 136                                 fl4->fl4_icmp_type = icmp[0];
 137                                 fl4->fl4_icmp_code = icmp[1];
 138                         }
 139                         break;
 140
 141                 case IPPROTO_ESP:
 142                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 143                                 __be32 *ehdr = (__be32 *)xprth;
 144
 145                                 fl4->fl4_ipsec_spi = ehdr[0];
 146                         }
 147                         break;
 148
 149                 case IPPROTO_AH:
 150                         if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
 151                                 __be32 *ah_hdr = (__be32 *)xprth;
 152
 153                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 154                         }
 155                         break;
 156
 157                 case IPPROTO_COMP:
 158                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 159                                 __be16 *ipcomp_hdr = (__be16 *)xprth;
 160
 161                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 162                         }
 163                         break;
 164
 165                 case IPPROTO_GRE:
 166                         if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
 167                                 __be16 *greflags = (__be16 *)xprth;
 168                                 __be32 *gre_hdr = (__be32 *)xprth;
 169
 170                                 if (greflags[0] & GRE_KEY) {
 171                                         if (greflags[0] & GRE_CSUM)
 172                                                 gre_hdr++;
 173                                         fl4->fl4_gre_key = gre_hdr[1];
 174                                 }
 175                         }
 176                         break;
 177
 178                 default:
 179                         fl4->fl4_ipsec_spi = 0;
 180                         break;
 181                 }
 182         }
 183         fl4->flowi4_proto = iph->protocol;
 184         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 185         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 186         fl4->flowi4_tos = iph->tos;
 187 }
 188
 189 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 190 {
 191         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 192
 193         xfrm4_policy_afinfo.garbage_collect(net);
 194         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 195 }
 196
 197 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 198                               struct sk_buff *skb, u32 mtu)
 199 {
 200         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 201         struct dst_entry *path = xdst->route;
 202
 203         path->ops->update_pmtu(path, sk, skb, mtu);
 204 }
 205
 206 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 207                            struct sk_buff *skb)
 208 {
 209         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 210         struct dst_entry *path = xdst->route;
 211
 212         path->ops->redirect(path, sk, skb);
 213 }
 214
 215 static void xfrm4_dst_destroy(struct dst_entry *dst)
 216 {
 217         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 218
 219         dst_destroy_metrics_generic(dst);
 220
 221         xfrm_dst_destroy(xdst);
 222 }
 223
 224 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 225                              int unregister)
 226 {
 227         if (!unregister)
 228                 return;
 229
 230         xfrm_dst_ifdown(dst, dev);
 231 }
 232
 233 static struct dst_ops xfrm4_dst_ops = {
 234         .family =               AF_INET,
 235         .protocol =             cpu_to_be16(ETH_P_IP),
 236         .gc =                   xfrm4_garbage_collect,
 237         .update_pmtu =          xfrm4_update_pmtu,
 238         .redirect =             xfrm4_redirect,
 239         .cow_metrics =          dst_cow_metrics_generic,
 240         .destroy =              xfrm4_dst_destroy,
 241         .ifdown =               xfrm4_dst_ifdown,
 242         .local_out =            __ip_local_out,
 243         .gc_thresh =            1024,
 244 };
 245
 246 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 247         .family =               AF_INET,
 248         .dst_ops =              &xfrm4_dst_ops,
 249         .dst_lookup =           xfrm4_dst_lookup,
 250         .get_saddr =            xfrm4_get_saddr,
 251         .decode_session =       _decode_session4,
 252         .get_tos =              xfrm4_get_tos,
 253         .init_path =            xfrm4_init_path,
 254         .fill_dst =             xfrm4_fill_dst,
 255         .blackhole_route =      ipv4_blackhole_route,
 256 };
 257
 258 #ifdef CONFIG_SYSCTL
 259 static struct ctl_table xfrm4_policy_table[] = {
 260         {
 261                 .procname       = "xfrm4_gc_thresh",
 262                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 263                 .maxlen         = sizeof(int),
 264                 .mode           = 0644,
 265                 .proc_handler   = proc_dointvec,
 266         },
 267         { }
 268 };
 269
 270 static int __net_init xfrm4_net_init(struct net *net)
 271 {
 272         struct ctl_table *table;
 273         struct ctl_table_header *hdr;
 274
 275         table = xfrm4_policy_table;
 276         if (!net_eq(net, &init_net)) {
 277                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 278                 if (!table)
 279                         goto err_alloc;
 280
 281                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 282         }
 283
 284         hdr = register_net_sysctl(net, "net/ipv4", table);
 285         if (!hdr)
 286                 goto err_reg;
 287
 288         net->ipv4.xfrm4_hdr = hdr;
 289         return 0;
 290
 291 err_reg:
 292         if (!net_eq(net, &init_net))
 293                 kfree(table);
 294 err_alloc:
 295         return -ENOMEM;
 296 }
 297
 298 static void __net_exit xfrm4_net_exit(struct net *net)
 299 {
 300         struct ctl_table *table;
 301
 302         if (net->ipv4.xfrm4_hdr == NULL)
 303                 return;
 304
 305         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 306         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 307         if (!net_eq(net, &init_net))
 308                 kfree(table);
 309 }
 310
 311 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 312         .init   = xfrm4_net_init,
 313         .exit   = xfrm4_net_exit,
 314 };
 315 #endif
 316
 317 static void __init xfrm4_policy_init(void)
 318 {
 319         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 320 }
 321
 322 void __init xfrm4_init(void)
 323 {
 324         dst_entries_init(&xfrm4_dst_ops);
 325
 326         xfrm4_state_init();
 327         xfrm4_policy_init();
 328 #ifdef CONFIG_SYSCTL
 329         register_pernet_subsys(&xfrm4_net_ops);
 330 #endif
 331 }
 332