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 flowi4 fl4 = {
  26                 .daddr = daddr->a4,
  27                 .flowi4_tos = tos,
  28         };
  29         struct rtable *rt;
  30
  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 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->u.ip4.flowi4_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         const struct flowi4 *fl4 = &fl->u.ip4;
  73
  74         rt->rt_key_dst = fl4->daddr;
  75         rt->rt_key_src = fl4->saddr;
  76         rt->rt_tos = fl4->flowi4_tos;
  77         rt->rt_route_iif = fl4->flowi4_iif;
  78         rt->rt_iif = fl4->flowi4_iif;
  79         rt->rt_oif = fl4->flowi4_oif;
  80         rt->rt_mark = fl4->flowi4_mark;
  81
  82         xdst->u.dst.dev = dev;
  83         dev_hold(dev);
  84
  85         xdst->u.rt.peer = rt->peer;
  86         if (rt->peer)
  87                 atomic_inc(&rt->peer->refcnt);
  88
  89         /* Sheit... I remember I did this right. Apparently,
  90          * it was magically lost, so this code needs audit */
  91         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
  92                                               RTCF_LOCAL);
  93         xdst->u.rt.rt_type = rt->rt_type;
  94         xdst->u.rt.rt_src = rt->rt_src;
  95         xdst->u.rt.rt_dst = rt->rt_dst;
  96         xdst->u.rt.rt_gateway = rt->rt_gateway;
  97         xdst->u.rt.rt_spec_dst = rt->rt_spec_dst;
  98
  99         return 0;
 100 }
 101
 102 static void
 103 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 104 {
 105         struct iphdr *iph = ip_hdr(skb);
 106         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 107         struct flowi4 *fl4 = &fl->u.ip4;
 108
 109         memset(fl4, 0, sizeof(struct flowi4));
 110         fl4->flowi4_mark = skb->mark;
 111
 112         if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) {
 113                 switch (iph->protocol) {
 114                 case IPPROTO_UDP:
 115                 case IPPROTO_UDPLITE:
 116                 case IPPROTO_TCP:
 117                 case IPPROTO_SCTP:
 118                 case IPPROTO_DCCP:
 119                         if (xprth + 4 < skb->data ||
 120                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 121                                 __be16 *ports = (__be16 *)xprth;
 122
 123                                 fl4->fl4_sport = ports[!!reverse];
 124                                 fl4->fl4_dport = ports[!reverse];
 125                         }
 126                         break;
 127
 128                 case IPPROTO_ICMP:
 129                         if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
 130                                 u8 *icmp = xprth;
 131
 132                                 fl4->fl4_icmp_type = icmp[0];
 133                                 fl4->fl4_icmp_code = icmp[1];
 134                         }
 135                         break;
 136
 137                 case IPPROTO_ESP:
 138                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 139                                 __be32 *ehdr = (__be32 *)xprth;
 140
 141                                 fl4->fl4_ipsec_spi = ehdr[0];
 142                         }
 143                         break;
 144
 145                 case IPPROTO_AH:
 146                         if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
 147                                 __be32 *ah_hdr = (__be32*)xprth;
 148
 149                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 150                         }
 151                         break;
 152
 153                 case IPPROTO_COMP:
 154                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 155                                 __be16 *ipcomp_hdr = (__be16 *)xprth;
 156
 157                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 158                         }
 159                         break;
 160
 161                 case IPPROTO_GRE:
 162                         if (pskb_may_pull(skb, xprth + 12 - skb->data)) {
 163                                 __be16 *greflags = (__be16 *)xprth;
 164                                 __be32 *gre_hdr = (__be32 *)xprth;
 165
 166                                 if (greflags[0] & GRE_KEY) {
 167                                         if (greflags[0] & GRE_CSUM)
 168                                                 gre_hdr++;
 169                                         fl4->fl4_gre_key = gre_hdr[1];
 170                                 }
 171                         }
 172                         break;
 173
 174                 default:
 175                         fl4->fl4_ipsec_spi = 0;
 176                         break;
 177                 }
 178         }
 179         fl4->flowi4_proto = iph->protocol;
 180         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 181         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 182         fl4->flowi4_tos = iph->tos;
 183 }
 184
 185 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 186 {
 187         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 188
 189         xfrm4_policy_afinfo.garbage_collect(net);
 190         return (dst_entries_get_slow(ops) > ops->gc_thresh * 2);
 191 }
 192
 193 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
 194 {
 195         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 196         struct dst_entry *path = xdst->route;
 197
 198         path->ops->update_pmtu(path, mtu);
 199 }
 200
 201 static void xfrm4_dst_destroy(struct dst_entry *dst)
 202 {
 203         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 204
 205         dst_destroy_metrics_generic(dst);
 206
 207         if (likely(xdst->u.rt.peer))
 208                 inet_putpeer(xdst->u.rt.peer);
 209
 210         xfrm_dst_destroy(xdst);
 211 }
 212
 213 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 214                              int unregister)
 215 {
 216         if (!unregister)
 217                 return;
 218
 219         xfrm_dst_ifdown(dst, dev);
 220 }
 221
 222 static struct dst_ops xfrm4_dst_ops = {
 223         .family =               AF_INET,
 224         .protocol =             cpu_to_be16(ETH_P_IP),
 225         .gc =                   xfrm4_garbage_collect,
 226         .update_pmtu =          xfrm4_update_pmtu,
 227         .cow_metrics =          dst_cow_metrics_generic,
 228         .destroy =              xfrm4_dst_destroy,
 229         .ifdown =               xfrm4_dst_ifdown,
 230         .local_out =            __ip_local_out,
 231         .gc_thresh =            1024,
 232 };
 233
 234 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 235         .family =               AF_INET,
 236         .dst_ops =              &xfrm4_dst_ops,
 237         .dst_lookup =           xfrm4_dst_lookup,
 238         .get_saddr =            xfrm4_get_saddr,
 239         .decode_session =       _decode_session4,
 240         .get_tos =              xfrm4_get_tos,
 241         .init_path =            xfrm4_init_path,
 242         .fill_dst =             xfrm4_fill_dst,
 243         .blackhole_route =      ipv4_blackhole_route,
 244 };
 245
 246 #ifdef CONFIG_SYSCTL
 247 static struct ctl_table xfrm4_policy_table[] = {
 248         {
 249                 .procname       = "xfrm4_gc_thresh",
 250                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 251                 .maxlen         = sizeof(int),
 252                 .mode           = 0644,
 253                 .proc_handler   = proc_dointvec,
 254         },
 255         { }
 256 };
 257
 258 static struct ctl_table_header *sysctl_hdr;
 259 #endif
 260
 261 static void __init xfrm4_policy_init(void)
 262 {
 263         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 264 }
 265
 266 static void __exit xfrm4_policy_fini(void)
 267 {
 268 #ifdef CONFIG_SYSCTL
 269         if (sysctl_hdr)
 270                 unregister_net_sysctl_table(sysctl_hdr);
 271 #endif
 272         xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
 273 }
 274
 275 void __init xfrm4_init(int rt_max_size)
 276 {
 277         /*
 278          * Select a default value for the gc_thresh based on the main route
 279          * table hash size.  It seems to me the worst case scenario is when
 280          * we have ipsec operating in transport mode, in which we create a
 281          * dst_entry per socket.  The xfrm gc algorithm starts trying to remove
 282          * entries at gc_thresh, and prevents new allocations as 2*gc_thresh
 283          * so lets set an initial xfrm gc_thresh value at the rt_max_size/2.
 284          * That will let us store an ipsec connection per route table entry,
 285          * and start cleaning when were 1/2 full
 286          */
 287         xfrm4_dst_ops.gc_thresh = rt_max_size/2;
 288         dst_entries_init(&xfrm4_dst_ops);
 289
 290         xfrm4_state_init();
 291         xfrm4_policy_init();
 292 #ifdef CONFIG_SYSCTL
 293         sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv4_ctl_path,
 294                                                 xfrm4_policy_table);
 295 #endif
 296 }
 297