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 xfrm_policy_afinfo xfrm4_policy_afinfo;
  19
  20 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos,
  21                                           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(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(struct net *net,
  47                            xfrm_address_t *saddr, xfrm_address_t *daddr)
  48 {
  49         struct dst_entry *dst;
  50         struct rtable *rt;
  51
  52         dst = xfrm4_dst_lookup(net, 0, NULL, daddr);
  53         if (IS_ERR(dst))
  54                 return -EHOSTUNREACH;
  55
  56         rt = (struct rtable *)dst;
  57         saddr->a4 = rt->rt_src;
  58         dst_release(dst);
  59         return 0;
  60 }
  61
  62 static struct dst_entry *
  63 __xfrm4_find_bundle(struct flowi *fl, struct xfrm_policy *policy)
  64 {
  65         struct dst_entry *dst;
  66
  67         read_lock_bh(&policy->lock);
  68         for (dst = policy->bundles; dst; dst = dst->next) {
  69                 struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
  70                 if (xdst->u.rt.fl.oif == fl->oif &&     /*XXX*/
  71                     xdst->u.rt.fl.fl4_dst == fl->fl4_dst &&
  72                     xdst->u.rt.fl.fl4_src == fl->fl4_src &&
  73                     xdst->u.rt.fl.fl4_tos == fl->fl4_tos &&
  74                     xfrm_bundle_ok(policy, xdst, fl, AF_INET, 0)) {
  75                         dst_clone(dst);
  76                         break;
  77                 }
  78         }
  79         read_unlock_bh(&policy->lock);
  80         return dst;
  81 }
  82
  83 static int xfrm4_get_tos(struct flowi *fl)
  84 {
  85         return fl->fl4_tos;
  86 }
  87
  88 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst,
  89                            int nfheader_len)
  90 {
  91         return 0;
  92 }
  93
  94 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev,
  95                           struct flowi *fl)
  96 {
  97         struct rtable *rt = (struct rtable *)xdst->route;
  98
  99         xdst->u.rt.fl = *fl;
 100
 101         xdst->u.dst.dev = dev;
 102         dev_hold(dev);
 103
 104         xdst->u.rt.idev = in_dev_get(dev);
 105         if (!xdst->u.rt.idev)
 106                 return -ENODEV;
 107
 108         xdst->u.rt.peer = rt->peer;
 109         if (rt->peer)
 110                 atomic_inc(&rt->peer->refcnt);
 111
 112         /* Sheit... I remember I did this right. Apparently,
 113          * it was magically lost, so this code needs audit */
 114         xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST |
 115                                               RTCF_LOCAL);
 116         xdst->u.rt.rt_type = rt->rt_type;
 117         xdst->u.rt.rt_src = rt->rt_src;
 118         xdst->u.rt.rt_dst = rt->rt_dst;
 119         xdst->u.rt.rt_gateway = rt->rt_gateway;
 120         xdst->u.rt.rt_spec_dst = rt->rt_spec_dst;
 121
 122         return 0;
 123 }
 124
 125 static void
 126 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 127 {
 128         struct iphdr *iph = ip_hdr(skb);
 129         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 130
 131         memset(fl, 0, sizeof(struct flowi));
 132         if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) {
 133                 switch (iph->protocol) {
 134                 case IPPROTO_UDP:
 135                 case IPPROTO_UDPLITE:
 136                 case IPPROTO_TCP:
 137                 case IPPROTO_SCTP:
 138                 case IPPROTO_DCCP:
 139                         if (xprth + 4 < skb->data ||
 140                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 141                                 __be16 *ports = (__be16 *)xprth;
 142
 143                                 fl->fl_ip_sport = ports[!!reverse];
 144                                 fl->fl_ip_dport = ports[!reverse];
 145                         }
 146                         break;
 147
 148                 case IPPROTO_ICMP:
 149                         if (pskb_may_pull(skb, xprth + 2 - skb->data)) {
 150                                 u8 *icmp = xprth;
 151
 152                                 fl->fl_icmp_type = icmp[0];
 153                                 fl->fl_icmp_code = icmp[1];
 154                         }
 155                         break;
 156
 157                 case IPPROTO_ESP:
 158                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 159                                 __be32 *ehdr = (__be32 *)xprth;
 160
 161                                 fl->fl_ipsec_spi = ehdr[0];
 162                         }
 163                         break;
 164
 165                 case IPPROTO_AH:
 166                         if (pskb_may_pull(skb, xprth + 8 - skb->data)) {
 167                                 __be32 *ah_hdr = (__be32*)xprth;
 168
 169                                 fl->fl_ipsec_spi = ah_hdr[1];
 170                         }
 171                         break;
 172
 173                 case IPPROTO_COMP:
 174                         if (pskb_may_pull(skb, xprth + 4 - skb->data)) {
 175                                 __be16 *ipcomp_hdr = (__be16 *)xprth;
 176
 177                                 fl->fl_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 178                         }
 179                         break;
 180                 default:
 181                         fl->fl_ipsec_spi = 0;
 182                         break;
 183                 }
 184         }
 185         fl->proto = iph->protocol;
 186         fl->fl4_dst = reverse ? iph->saddr : iph->daddr;
 187         fl->fl4_src = reverse ? iph->daddr : iph->saddr;
 188         fl->fl4_tos = iph->tos;
 189 }
 190
 191 static inline int xfrm4_garbage_collect(struct dst_ops *ops)
 192 {
 193         struct net *net = container_of(ops, struct net, xfrm.xfrm4_dst_ops);
 194
 195         xfrm4_policy_afinfo.garbage_collect(net);
 196         return (atomic_read(&ops->entries) > ops->gc_thresh * 2);
 197 }
 198
 199 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu)
 200 {
 201         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 202         struct dst_entry *path = xdst->route;
 203
 204         path->ops->update_pmtu(path, mtu);
 205 }
 206
 207 static void xfrm4_dst_destroy(struct dst_entry *dst)
 208 {
 209         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 210
 211         if (likely(xdst->u.rt.idev))
 212                 in_dev_put(xdst->u.rt.idev);
 213         if (likely(xdst->u.rt.peer))
 214                 inet_putpeer(xdst->u.rt.peer);
 215         xfrm_dst_destroy(xdst);
 216 }
 217
 218 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 219                              int unregister)
 220 {
 221         struct xfrm_dst *xdst;
 222
 223         if (!unregister)
 224                 return;
 225
 226         xdst = (struct xfrm_dst *)dst;
 227         if (xdst->u.rt.idev->dev == dev) {
 228                 struct in_device *loopback_idev =
 229                         in_dev_get(dev_net(dev)->loopback_dev);
 230                 BUG_ON(!loopback_idev);
 231
 232                 do {
 233                         in_dev_put(xdst->u.rt.idev);
 234                         xdst->u.rt.idev = loopback_idev;
 235                         in_dev_hold(loopback_idev);
 236                         xdst = (struct xfrm_dst *)xdst->u.dst.child;
 237                 } while (xdst->u.dst.xfrm);
 238
 239                 __in_dev_put(loopback_idev);
 240         }
 241
 242         xfrm_dst_ifdown(dst, dev);
 243 }
 244
 245 static struct dst_ops xfrm4_dst_ops = {
 246         .family =               AF_INET,
 247         .protocol =             cpu_to_be16(ETH_P_IP),
 248         .gc =                   xfrm4_garbage_collect,
 249         .update_pmtu =          xfrm4_update_pmtu,
 250         .destroy =              xfrm4_dst_destroy,
 251         .ifdown =               xfrm4_dst_ifdown,
 252         .local_out =            __ip_local_out,
 253         .gc_thresh =            1024,
 254         .entries =              ATOMIC_INIT(0),
 255 };
 256
 257 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 258         .family =               AF_INET,
 259         .dst_ops =              &xfrm4_dst_ops,
 260         .dst_lookup =           xfrm4_dst_lookup,
 261         .get_saddr =            xfrm4_get_saddr,
 262         .find_bundle =          __xfrm4_find_bundle,
 263         .decode_session =       _decode_session4,
 264         .get_tos =              xfrm4_get_tos,
 265         .init_path =            xfrm4_init_path,
 266         .fill_dst =             xfrm4_fill_dst,
 267 };
 268
 269 #ifdef CONFIG_SYSCTL
 270 static struct ctl_table xfrm4_policy_table[] = {
 271         {
 272                 .procname       = "xfrm4_gc_thresh",
 273                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 274                 .maxlen         = sizeof(int),
 275                 .mode           = 0644,
 276                 .proc_handler   = proc_dointvec,
 277         },
 278         { }
 279 };
 280
 281 static struct ctl_table_header *sysctl_hdr;
 282 #endif
 283
 284 static void __init xfrm4_policy_init(void)
 285 {
 286         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo);
 287 }
 288
 289 static void __exit xfrm4_policy_fini(void)
 290 {
 291 #ifdef CONFIG_SYSCTL
 292         if (sysctl_hdr)
 293                 unregister_net_sysctl_table(sysctl_hdr);
 294 #endif
 295         xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo);
 296 }
 297
 298 void __init xfrm4_init(int rt_max_size)
 299 {
 300         /*
 301          * Select a default value for the gc_thresh based on the main route
 302          * table hash size.  It seems to me the worst case scenario is when
 303          * we have ipsec operating in transport mode, in which we create a
 304          * dst_entry per socket.  The xfrm gc algorithm starts trying to remove
 305          * entries at gc_thresh, and prevents new allocations as 2*gc_thresh
 306          * so lets set an initial xfrm gc_thresh value at the rt_max_size/2.
 307          * That will let us store an ipsec connection per route table entry,
 308          * and start cleaning when were 1/2 full
 309          */
 310         xfrm4_dst_ops.gc_thresh = rt_max_size/2;
 311
 312         xfrm4_state_init();
 313         xfrm4_policy_init();
 314 #ifdef CONFIG_SYSCTL
 315         sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv4_ctl_path,
 316                                                 xfrm4_policy_table);
 317 #endif
 318 }
 319