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, int tos,
  22                                           const xfrm_address_t *saddr,
  23                                           const xfrm_address_t *daddr)
  24 {
  25         struct flowi fl = {
  26                 .fl4_dst = daddr->a4,
  27                 .fl4_tos = tos,
  28         };
  29         struct rtable *rt;
  30
  31         if (saddr)
  32                 fl.fl4_src = saddr->a4;
  33
  34         rt = __ip_route_output_key(net, &fl);
  35         if (!IS_ERR(rt))
  36                 return &rt->dst;
  37
  38         return ERR_CAST(rt);
  39 }
  40
  41 static int xfrm4_get_saddr(struct net *net,
  42                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  43 {
  44         struct dst_entry *dst;
  45         struct rtable *rt;
  46
  47         dst = xfrm4_dst_lookup(net, 0, NULL, daddr);
  48         if (IS_ERR(dst))
  49                 return -EHOSTUNREACH;
  50
  51         rt = (struct rtable *)dst;
  52         saddr->a4 = rt->rt_src;
  53         dst_release(dst);
  54         return 0;
  55 }
  56
  57 static int xfrm4_get_tos(const struct flowi *fl)
  58 {
  59         return IPTOS_RT_MASK & fl->fl4_tos; /* Strip ECN bits */
  60 }
  61
  62 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  63                            int nfheader_len)
  64 {
  65         return 0;
  66 }
  67
  68 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  69                           const struct flowi *fl)
  70 {
  71         struct rtable *rt = (struct rtable *)xdst->route;
  72
  73         rt->rt_key_dst = fl->fl4_dst;
  74         rt->rt_key_src = fl->fl4_src;
  75         rt->rt_tos = fl->fl4_tos;
  76         rt->rt_iif = fl->flowi_iif;
  77         rt->rt_oif = fl->flowi_oif;
  78         rt->rt_mark = fl->flowi_mark;
  79
  80         xdst->u.dst.dev = dev;
  81         dev_hold(dev);
  82
  83         xdst->u.rt.peer = rt->peer;
  84         if (rt->peer)
  85                 atomic_inc(&rt->peer->refcnt);
  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_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
  90                                               RTCF_LOCAL);
  91         xdst->u.rt.rt_type = rt->rt_type;
  92         xdst->u.rt.rt_src = rt->rt_src;
  93         xdst->u.rt.rt_dst = rt->rt_dst;
  94         xdst->u.rt.rt_gateway = rt->rt_gateway;
  95         xdst->u.rt.rt_spec_dst = rt->rt_spec_dst;
  96
  97         return 0;
  98 }
  99
 100 static void
 101 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 102 {
 103         struct iphdr *iph = ip_hdr(skb);
 104         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 105
 106         memset(fl, 0, sizeof(struct flowi));
 107         fl->flowi_mark = skb->mark;
 108
 109         if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) {
 110                 switch (iph->protocol) {
 111                 case IPPROTO_UDP:
 112                 case IPPROTO_UDPLITE:
 113                 case IPPROTO_TCP:
 114                 case IPPROTO_SCTP:
 115                 case IPPROTO_DCCP:
 116                         if (xprth + 4 < skb->data ||
 117                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 118                                 __be16 *ports = (__be16 *)xprth;
 119
 120                                 fl->fl_ip_sport = ports[!!reverse];
 121                                 fl->fl_ip_dport = ports[!reverse];
 122                         }
 123                         break;
 124
 125                 case IPPROTO_ICMP:
 126                         if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
 127                                 u8 *icmp = xprth;
 128
 129                                 fl->fl_icmp_type = icmp[0];
 130                                 fl->fl_icmp_code = icmp[1];
 131                         }
 132                         break;
 133
 134                 case IPPROTO_ESP:
 135                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 136                                 __be32 *ehdr = (__be32 *)xprth;
 137
 138                                 fl->fl_ipsec_spi = ehdr[0];
 139                         }
 140                         break;
 141
 142                 case IPPROTO_AH:
 143                         if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
 144                                 __be32 *ah_hdr = (__be32*)xprth;
 145
 146                                 fl->fl_ipsec_spi = ah_hdr[1];
 147                         }
 148                         break;
 149
 150                 case IPPROTO_COMP:
 151                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 152                                 __be16 *ipcomp_hdr = (__be16 *)xprth;
 153
 154                                 fl->fl_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 155                         }
 156                         break;
 157
 158                 case IPPROTO_GRE:
 159                         if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
 160                                 __be16 *greflags = (__be16 *)xprth;
 161                                 __be32 *gre_hdr = (__be32 *)xprth;
 162
 163                                 if (greflags[0] & GRE_KEY) {
 164                                         if (greflags[0] & GRE_CSUM)
 165                                                 gre_hdr++;
 166                                         fl->fl_gre_key = gre_hdr[1];
 167                                 }
 168                         }
 169                         break;
 170
 171                 default:
 172                         fl->fl_ipsec_spi = 0;
 173                         break;
 174                 }
 175         }
 176         fl->flowi_proto = iph->protocol;
 177         fl->fl4_dst = reverse ? iph->saddr : iph->daddr;
 178         fl->fl4_src = reverse ? iph->daddr : iph->saddr;
 179         fl->fl4_tos = iph->tos;
 180 }
 181
 182 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 183 {
 184         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 185
 186         xfrm4_policy_afinfo.garbage_collect(net);
 187         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 188 }
 189
 190 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
 191 {
 192         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 193         struct dst_entry *path = xdst->route;
 194
 195         path->ops->update_pmtu(path, mtu);
 196 }
 197
 198 static void xfrm4_dst_destroy(struct dst_entry *dst)
 199 {
 200         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 201
 202         dst_destroy_metrics_generic(dst);
 203
 204         if (likely(xdst->u.rt.peer))
 205                 inet_putpeer(xdst->u.rt.peer);
 206
 207         xfrm_dst_destroy(xdst);
 208 }
 209
 210 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 211                              int unregister)
 212 {
 213         if (!unregister)
 214                 return;
 215
 216         xfrm_dst_ifdown(dst, dev);
 217 }
 218
 219 static struct dst_ops xfrm4_dst_ops = {
 220         .family =               AF_INET,
 221         .protocol =             cpu_to_be16(ETH_P_IP),
 222         .gc =                   xfrm4_garbage_collect,
 223         .update_pmtu =          xfrm4_update_pmtu,
 224         .cow_metrics =          dst_cow_metrics_generic,
 225         .destroy =              xfrm4_dst_destroy,
 226         .ifdown =               xfrm4_dst_ifdown,
 227         .local_out =            __ip_local_out,
 228         .gc_thresh =            1024,
 229 };
 230
 231 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 232         .family =               AF_INET,
 233         .dst_ops =              &xfrm4_dst_ops,
 234         .dst_lookup =           xfrm4_dst_lookup,
 235         .get_saddr =            xfrm4_get_saddr,
 236         .decode_session =       _decode_session4,
 237         .get_tos =              xfrm4_get_tos,
 238         .init_path =            xfrm4_init_path,
 239         .fill_dst =             xfrm4_fill_dst,
 240         .blackhole_route =      ipv4_blackhole_route,
 241 };
 242
 243 #ifdef CONFIG_SYSCTL
 244 static struct ctl_table xfrm4_policy_table[] = {
 245         {
 246                 .procname       = "xfrm4_gc_thresh",
 247                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 248                 .maxlen         = sizeof(int),
 249                 .mode           = 0644,
 250                 .proc_handler   = proc_dointvec,
 251         },
 252         { }
 253 };
 254
 255 static struct ctl_table_header *sysctl_hdr;
 256 #endif
 257
 258 static void __init xfrm4_policy_init(void)
 259 {
 260         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 261 }
 262
 263 static void __exit xfrm4_policy_fini(void)
 264 {
 265 #ifdef CONFIG_SYSCTL
 266         if (sysctl_hdr)
 267                 unregister_net_sysctl_table(sysctl_hdr);
 268 #endif
 269         xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
 270 }
 271
 272 void __init xfrm4_init(int rt_max_size)
 273 {
 274         /*
 275          * Select a default value for the gc_thresh based on the main route
 276          * table hash size.  It seems to me the worst case scenario is when
 277          * we have ipsec operating in transport mode, in which we create a
 278          * dst_entry per socket.  The xfrm gc algorithm starts trying to remove
 279          * entries at gc_thresh, and prevents new allocations as 2*gc_thresh
 280          * so lets set an initial xfrm gc_thresh value at the rt_max_size/2.
 281          * That will let us store an ipsec connection per route table entry,
 282          * and start cleaning when were 1/2 full
 283          */
 284         xfrm4_dst_ops.gc_thresh = rt_max_size/2;
 285         dst_entries_init(&xfrm4_dst_ops);
 286
 287         xfrm4_state_init();
 288         xfrm4_policy_init();
 289 #ifdef CONFIG_SYSCTL
 290         sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv4_ctl_path,
 291                                                 xfrm4_policy_table);
 292 #endif
 293 }
 294