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/compiler.h> 12 #include <linux/config.h> 13 #include <linux/inetdevice.h> 14 #include <net/xfrm.h> 15 #include <net/ip.h> 16 17 static struct dst_ops xfrm4_dst_ops; 18 static struct xfrm_policy_afinfo xfrm4_policy_afinfo; 19 20 static int xfrm4_dst_lookup(struct xfrm_dst **dst, struct flowi *fl) 21 { 22 return __ip_route_output_key((struct rtable**)dst, fl); 23 } 24 25 static struct dst_entry * 26 __xfrm4_find_bundle(struct flowi *fl, struct xfrm_policy *policy) 27 { 28 struct dst_entry *dst; 29 30 read_lock_bh(&policy->lock); 31 for (dst = policy->bundles; dst; dst = dst->next) { 32 struct xfrm_dst *xdst = (struct xfrm_dst*)dst; 33 if (xdst->u.rt.fl.oif == fl->oif && /*XXX*/ 34 xdst->u.rt.fl.fl4_dst == fl->fl4_dst && 35 xdst->u.rt.fl.fl4_src == fl->fl4_src && 36 xdst->u.rt.fl.fl4_tos == fl->fl4_tos && 37 xfrm_bundle_ok(xdst, fl, AF_INET)) { 38 dst_clone(dst); 39 break; 40 } 41 } 42 read_unlock_bh(&policy->lock); 43 return dst; 44 } 45 46 /* Allocate chain of dst_entry's, attach known xfrm's, calculate 47 * all the metrics... Shortly, bundle a bundle. 48 */ 49 50 static int 51 __xfrm4_bundle_create(struct xfrm_policy *policy, struct xfrm_state **xfrm, int nx, 52 struct flowi *fl, struct dst_entry **dst_p) 53 { 54 struct dst_entry *dst, *dst_prev; 55 struct rtable *rt0 = (struct rtable*)(*dst_p); 56 struct rtable *rt = rt0; 57 u32 remote = fl->fl4_dst; 58 u32 local = fl->fl4_src; 59 struct flowi fl_tunnel = { 60 .nl_u = { 61 .ip4_u = { 62 .saddr = local, 63 .daddr = remote, 64 .tos = fl->fl4_tos 65 } 66 } 67 }; 68 int i; 69 int err; 70 int header_len = 0; 71 int trailer_len = 0; 72 73 dst = dst_prev = NULL; 74 dst_hold(&rt->u.dst); 75 76 for (i = 0; i < nx; i++) { 77 struct dst_entry *dst1 = dst_alloc(&xfrm4_dst_ops); 78 struct xfrm_dst *xdst; 79 int tunnel = 0; 80 81 if (unlikely(dst1 == NULL)) { 82 err = -ENOBUFS; 83 dst_release(&rt->u.dst); 84 goto error; 85 } 86 87 if (!dst) 88 dst = dst1; 89 else { 90 dst_prev->child = dst1; 91 dst1->flags |= DST_NOHASH; 92 dst_clone(dst1); 93 } 94 95 xdst = (struct xfrm_dst *)dst1; 96 xdst->route = &rt->u.dst; 97 98 dst1->next = dst_prev; 99 dst_prev = dst1; 100 if (xfrm[i]->props.mode) { 101 remote = xfrm[i]->id.daddr.a4; 102 local = xfrm[i]->props.saddr.a4; 103 tunnel = 1; 104 } 105 header_len += xfrm[i]->props.header_len; 106 trailer_len += xfrm[i]->props.trailer_len; 107 108 if (tunnel) { 109 fl_tunnel.fl4_src = local; 110 fl_tunnel.fl4_dst = remote; 111 err = xfrm_dst_lookup((struct xfrm_dst **)&rt, 112 &fl_tunnel, AF_INET); 113 if (err) 114 goto error; 115 } else 116 dst_hold(&rt->u.dst); 117 } 118 119 dst_prev->child = &rt->u.dst; 120 dst->path = &rt->u.dst; 121 122 *dst_p = dst; 123 dst = dst_prev; 124 125 dst_prev = *dst_p; 126 i = 0; 127 for (; dst_prev != &rt->u.dst; dst_prev = dst_prev->child) { 128 struct xfrm_dst *x = (struct xfrm_dst*)dst_prev; 129 x->u.rt.fl = *fl; 130 131 dst_prev->xfrm = xfrm[i++]; 132 dst_prev->dev = rt->u.dst.dev; 133 if (rt->u.dst.dev) 134 dev_hold(rt->u.dst.dev); 135 dst_prev->obsolete = -1; 136 dst_prev->flags |= DST_HOST; 137 dst_prev->lastuse = jiffies; 138 dst_prev->header_len = header_len; 139 dst_prev->trailer_len = trailer_len; 140 memcpy(&dst_prev->metrics, &x->route->metrics, sizeof(dst_prev->metrics)); 141 142 /* Copy neighbout for reachability confirmation */ 143 dst_prev->neighbour = neigh_clone(rt->u.dst.neighbour); 144 dst_prev->input = rt->u.dst.input; 145 dst_prev->output = xfrm4_output; 146 if (rt->peer) 147 atomic_inc(&rt->peer->refcnt); 148 x->u.rt.peer = rt->peer; 149 /* Sheit... I remember I did this right. Apparently, 150 * it was magically lost, so this code needs audit */ 151 x->u.rt.rt_flags = rt0->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST|RTCF_LOCAL); 152 x->u.rt.rt_type = rt->rt_type; 153 x->u.rt.rt_src = rt0->rt_src; 154 x->u.rt.rt_dst = rt0->rt_dst; 155 x->u.rt.rt_gateway = rt->rt_gateway; 156 x->u.rt.rt_spec_dst = rt0->rt_spec_dst; 157 x->u.rt.idev = rt0->idev; 158 in_dev_hold(rt0->idev); 159 header_len -= x->u.dst.xfrm->props.header_len; 160 trailer_len -= x->u.dst.xfrm->props.trailer_len; 161 } 162 163 xfrm_init_pmtu(dst); 164 return 0; 165 166 error: 167 if (dst) 168 dst_free(dst); 169 return err; 170 } 171 172 static void 173 _decode_session4(struct sk_buff *skb, struct flowi *fl) 174 { 175 struct iphdr *iph = skb->nh.iph; 176 u8 *xprth = skb->nh.raw + iph->ihl*4; 177 178 memset(fl, 0, sizeof(struct flowi)); 179 if (!(iph->frag_off & htons(IP_MF | IP_OFFSET))) { 180 switch (iph->protocol) { 181 case IPPROTO_UDP: 182 case IPPROTO_TCP: 183 case IPPROTO_SCTP: 184 case IPPROTO_DCCP: 185 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 186 u16 *ports = (u16 *)xprth; 187 188 fl->fl_ip_sport = ports[0]; 189 fl->fl_ip_dport = ports[1]; 190 } 191 break; 192 193 case IPPROTO_ICMP: 194 if (pskb_may_pull(skb, xprth + 2 - skb->data)) { 195 u8 *icmp = xprth; 196 197 fl->fl_icmp_type = icmp[0]; 198 fl->fl_icmp_code = icmp[1]; 199 } 200 break; 201 202 case IPPROTO_ESP: 203 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 204 u32 *ehdr = (u32 *)xprth; 205 206 fl->fl_ipsec_spi = ehdr[0]; 207 } 208 break; 209 210 case IPPROTO_AH: 211 if (pskb_may_pull(skb, xprth + 8 - skb->data)) { 212 u32 *ah_hdr = (u32*)xprth; 213 214 fl->fl_ipsec_spi = ah_hdr[1]; 215 } 216 break; 217 218 case IPPROTO_COMP: 219 if (pskb_may_pull(skb, xprth + 4 - skb->data)) { 220 u16 *ipcomp_hdr = (u16 *)xprth; 221 222 fl->fl_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); 223 } 224 break; 225 default: 226 fl->fl_ipsec_spi = 0; 227 break; 228 }; 229 } 230 fl->proto = iph->protocol; 231 fl->fl4_dst = iph->daddr; 232 fl->fl4_src = iph->saddr; 233 fl->fl4_tos = iph->tos; 234 } 235 236 static inline int xfrm4_garbage_collect(void) 237 { 238 xfrm4_policy_afinfo.garbage_collect(); 239 return (atomic_read(&xfrm4_dst_ops.entries) > xfrm4_dst_ops.gc_thresh*2); 240 } 241 242 static void xfrm4_update_pmtu(struct dst_entry *dst, u32 mtu) 243 { 244 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 245 struct dst_entry *path = xdst->route; 246 247 path->ops->update_pmtu(path, mtu); 248 } 249 250 static void xfrm4_dst_destroy(struct dst_entry *dst) 251 { 252 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 253 254 if (likely(xdst->u.rt.idev)) 255 in_dev_put(xdst->u.rt.idev); 256 xfrm_dst_destroy(xdst); 257 } 258 259 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 260 int unregister) 261 { 262 struct xfrm_dst *xdst; 263 264 if (!unregister) 265 return; 266 267 xdst = (struct xfrm_dst *)dst; 268 if (xdst->u.rt.idev->dev == dev) { 269 struct in_device *loopback_idev = in_dev_get(&loopback_dev); 270 BUG_ON(!loopback_idev); 271 272 do { 273 in_dev_put(xdst->u.rt.idev); 274 xdst->u.rt.idev = loopback_idev; 275 in_dev_hold(loopback_idev); 276 xdst = (struct xfrm_dst *)xdst->u.dst.child; 277 } while (xdst->u.dst.xfrm); 278 279 __in_dev_put(loopback_idev); 280 } 281 282 xfrm_dst_ifdown(dst, dev); 283 } 284 285 static struct dst_ops xfrm4_dst_ops = { 286 .family = AF_INET, 287 .protocol = __constant_htons(ETH_P_IP), 288 .gc = xfrm4_garbage_collect, 289 .update_pmtu = xfrm4_update_pmtu, 290 .destroy = xfrm4_dst_destroy, 291 .ifdown = xfrm4_dst_ifdown, 292 .gc_thresh = 1024, 293 .entry_size = sizeof(struct xfrm_dst), 294 }; 295 296 static struct xfrm_policy_afinfo xfrm4_policy_afinfo = { 297 .family = AF_INET, 298 .dst_ops = &xfrm4_dst_ops, 299 .dst_lookup = xfrm4_dst_lookup, 300 .find_bundle = __xfrm4_find_bundle, 301 .bundle_create = __xfrm4_bundle_create, 302 .decode_session = _decode_session4, 303 }; 304 305 static void __init xfrm4_policy_init(void) 306 { 307 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo); 308 } 309 310 static void __exit xfrm4_policy_fini(void) 311 { 312 xfrm_policy_unregister_afinfo(&xfrm4_policy_afinfo); 313 } 314 315 void __init xfrm4_init(void) 316 { 317 xfrm4_state_init(); 318 xfrm4_policy_init(); 319 } 320 321