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 <net/dst.h>
  15 #include <net/xfrm.h>
  16 #include <net/ip.h>
  17
  18 static struct dst_ops xfrm4_dst_ops;
  19 static struct xfrm_policy_afinfo xfrm4_policy_afinfo;
  20
  21 static struct dst_entry *xfrm4_dst_lookup(int tos, xfrm_address_t *saddr,
  22                                           xfrm_address_t *daddr)
  23 {
  24         struct flowi fl = {
  25                 .nl_u = {
  26                         .ip4_u = {
  27                                 .tos = tos,
  28                                 .daddr = daddr->a4,
  29                         },
  30                 },
  31         };
  32         struct dst_entry *dst;
  33         struct rtable *rt;
  34         int err;
  35
  36         if (saddr)
  37                 fl.fl4_src = saddr->a4;
  38
  39         err = __ip_route_output_key(&init_net, &rt, &fl);
  40         dst = &rt->u.dst;
  41         if (err)
  42                 dst = ERR_PTR(err);
  43         return dst;
  44 }
  45
  46 static int xfrm4_get_saddr(xfrm_address_t *saddr, xfrm_address_t *daddr)
  47 {
  48         struct dst_entry *dst;
  49         struct rtable *rt;
  50
  51         dst = xfrm4_dst_lookup(0, NULL, daddr);
  52         if (IS_ERR(dst))
  53                 return -EHOSTUNREACH;
  54
  55         rt = (struct rtable *)dst;
  56         saddr->a4 = rt->rt_src;
  57         dst_release(dst);
  58         return 0;
  59 }
  60
  61 static struct dst_entry *
  62 __xfrm4_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
  63 {
  64         struct dst_entry *dst;
  65
  66         read_lock_bh(&policy->lock);
  67         for (dst = policy->bundles; dst; dst = dst->next) {
  68                 struct xfrm_dst *xdst = (struct xfrm_dst*)dst;
  69                 if (xdst->u.rt.fl.oif == fl->oif &&     /*XXX*/
  70                     xdst->u.rt.fl.fl4_dst == fl->fl4_dst &&
  71                     xdst->u.rt.fl.fl4_src == fl->fl4_src &&
  72                     xdst->u.rt.fl.fl4_tos == fl->fl4_tos &&
  73                     xfrm_bundle_ok(policy, xdst, fl, AF_INET, 0)) {
  74                         dst_clone(dst);
  75                         break;
  76                 }
  77         }
  78         read_unlock_bh(&policy->lock);
  79         return dst;
  80 }
  81
  82 static int xfrm4_get_tos(struct flowi *fl)
  83 {
  84         return fl->fl4_tos;
  85 }
  86
  87 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  88                            int nfheader_len)
  89 {
  90         return 0;
  91 }
  92
  93 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev)
  94 {
  95         struct rtable *rt = (struct rtable *)xdst->route;
  96
  97         xdst->u.rt.fl = rt->fl;
  98
  99         xdst->u.dst.dev = dev;
 100         dev_hold(dev);
 101
 102         xdst->u.rt.idev = in_dev_get(dev);
 103         if (!xdst->u.rt.idev)
 104                 return -ENODEV;
 105
 106         xdst->u.rt.peer = rt->peer;
 107         if (rt->peer)
 108                 atomic_inc(&rt->peer->refcnt);
 109
 110         /* Sheit... I remember I did this right. Apparently,
 111          * it was magically lost, so this code needs audit */
 112         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
 113                                               RTCF_LOCAL);
 114         xdst->u.rt.rt_type = rt->rt_type;
 115         xdst->u.rt.rt_src = rt->rt_src;
 116         xdst->u.rt.rt_dst = rt->rt_dst;
 117         xdst->u.rt.rt_gateway = rt->rt_gateway;
 118         xdst->u.rt.rt_spec_dst = rt->rt_spec_dst;
 119
 120         return 0;
 121 }
 122
 123 static void
 124 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 125 {
 126         struct iphdr *iph = ip_hdr(skb);
 127         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 128
 129         memset(fl, 0, sizeof(struct flowi));
 130         if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) {
 131                 switch (iph->protocol) {
 132                 case IPPROTO_UDP:
 133                 case IPPROTO_UDPLITE:
 134                 case IPPROTO_TCP:
 135                 case IPPROTO_SCTP:
 136                 case IPPROTO_DCCP:
 137                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 138                                 __be16 *ports = (__be16 *)xprth;
 139
 140                                 fl->fl_ip_sport = ports[!!reverse];
 141                                 fl->fl_ip_dport = ports[!reverse];
 142                         }
 143                         break;
 144
 145                 case IPPROTO_ICMP:
 146                         if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
 147                                 u8 *icmp = xprth;
 148
 149                                 fl->fl_icmp_type = icmp[0];
 150                                 fl->fl_icmp_code = icmp[1];
 151                         }
 152                         break;
 153
 154                 case IPPROTO_ESP:
 155                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 156                                 __be32 *ehdr = (__be32 *)xprth;
 157
 158                                 fl->fl_ipsec_spi = ehdr[0];
 159                         }
 160                         break;
 161
 162                 case IPPROTO_AH:
 163                         if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
 164                                 __be32 *ah_hdr = (__be32*)xprth;
 165
 166                                 fl->fl_ipsec_spi = ah_hdr[1];
 167                         }
 168                         break;
 169
 170                 case IPPROTO_COMP:
 171                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 172                                 __be16 *ipcomp_hdr = (__be16 *)xprth;
 173
 174                                 fl->fl_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 175                         }
 176                         break;
 177                 default:
 178                         fl->fl_ipsec_spi = 0;
 179                         break;
 180                 }
 181         }
 182         fl->proto = iph->protocol;
 183         fl->fl4_dst = reverse ? iph->saddr : iph->daddr;
 184         fl->fl4_src = reverse ? iph->daddr : iph->saddr;
 185         fl->fl4_tos = iph->tos;
 186 }
 187
 188 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 189 {
 190         xfrm4_policy_afinfo.garbage_collect();
 191         return (atomic_read(&xfrm4_dst_ops.entries) > xfrm4_dst_ops.gc_thresh*2);
 192 }
 193
 194 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
 195 {
 196         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 197         struct dst_entry *path = xdst->route;
 198
 199         path->ops->update_pmtu(path, mtu);
 200 }
 201
 202 static void xfrm4_dst_destroy(struct dst_entry *dst)
 203 {
 204         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 205
 206         if (likely(xdst->u.rt.idev))
 207                 in_dev_put(xdst->u.rt.idev);
 208         if (likely(xdst->u.rt.peer))
 209                 inet_putpeer(xdst->u.rt.peer);
 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         struct xfrm_dst *xdst;
 217
 218         if (!unregister)
 219                 return;
 220
 221         xdst = (struct xfrm_dst *)dst;
 222         if (xdst->u.rt.idev->dev == dev) {
 223                 struct in_device *loopback_idev =
 224                         in_dev_get(dev_net(dev)->loopback_dev);
 225                 BUG_ON(!loopback_idev);
 226
 227                 do {
 228                         in_dev_put(xdst->u.rt.idev);
 229                         xdst->u.rt.idev = loopback_idev;
 230                         in_dev_hold(loopback_idev);
 231                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 232                 } while (xdst->u.dst.xfrm);
 233
 234                 __in_dev_put(loopback_idev);
 235         }
 236
 237         xfrm_dst_ifdown(dst, dev);
 238 }
 239
 240 static struct dst_ops xfrm4_dst_ops = {
 241         .family =               AF_INET,
 242         .protocol =             __constant_htons(ETH_P_IP),
 243         .gc =                   xfrm4_garbage_collect,
 244         .update_pmtu =          xfrm4_update_pmtu,
 245         .destroy =              xfrm4_dst_destroy,
 246         .ifdown =               xfrm4_dst_ifdown,
 247         .local_out =            __ip_local_out,
 248         .gc_thresh =            1024,
 249         .entry_size =           sizeof(struct xfrm_dst),
 250         .entries =              ATOMIC_INIT(0),
 251 };
 252
 253 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 254         .family =               AF_INET,
 255         .dst_ops =              &xfrm4_dst_ops,
 256         .dst_lookup =           xfrm4_dst_lookup,
 257         .get_saddr =            xfrm4_get_saddr,
 258         .find_bundle =          __xfrm4_find_bundle,
 259         .decode_session =       _decode_session4,
 260         .get_tos =              xfrm4_get_tos,
 261         .init_path =            xfrm4_init_path,
 262         .fill_dst =             xfrm4_fill_dst,
 263 };
 264
 265 static void __init xfrm4_policy_init(void)
 266 {
 267         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 268 }
 269
 270 static void __exit xfrm4_policy_fini(void)
 271 {
 272         xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
 273 }
 274
 275 void __init xfrm4_init(void)
 276 {
 277         xfrm4_state_init();
 278         xfrm4_policy_init();
 279 }
 280