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