1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * xfrm4_policy.c 4 * 5 * Changes: 6 * Kazunori MIYAZAWA @USAGI 7 * YOSHIFUJI Hideaki @USAGI 8 * Split up af-specific portion 9 * 10 */ 11 12 #include <linux/err.h> 13 #include <linux/kernel.h> 14 #include <linux/inetdevice.h> 15 #include <linux/if_tunnel.h> 16 #include <net/dst.h> 17 #include <net/xfrm.h> 18 #include <net/ip.h> 19 #include <net/l3mdev.h> 20 21 static struct dst_entry *__xfrm4_dst_lookup(struct net *net, struct flowi4 *fl4, 22 int tos, int oif, 23 const xfrm_address_t *saddr, 24 const xfrm_address_t *daddr, 25 u32 mark) 26 { 27 struct rtable *rt; 28 29 memset(fl4, 0, sizeof(*fl4)); 30 fl4->daddr = daddr->a4; 31 fl4->flowi4_tos = tos; 32 fl4->flowi4_oif = l3mdev_master_ifindex_by_index(net, oif); 33 fl4->flowi4_mark = mark; 34 if (saddr) 35 fl4->saddr = saddr->a4; 36 37 fl4->flowi4_flags = FLOWI_FLAG_SKIP_NH_OIF; 38 39 rt = __ip_route_output_key(net, fl4); 40 if (!IS_ERR(rt)) 41 return &rt->dst; 42 43 return ERR_CAST(rt); 44 } 45 46 static struct dst_entry *xfrm4_dst_lookup(struct net *net, int tos, int oif, 47 const xfrm_address_t *saddr, 48 const xfrm_address_t *daddr, 49 u32 mark) 50 { 51 struct flowi4 fl4; 52 53 return __xfrm4_dst_lookup(net, &fl4, tos, oif, saddr, daddr, mark); 54 } 55 56 static int xfrm4_get_saddr(struct net *net, int oif, 57 xfrm_address_t *saddr, xfrm_address_t *daddr, 58 u32 mark) 59 { 60 struct dst_entry *dst; 61 struct flowi4 fl4; 62 63 dst = __xfrm4_dst_lookup(net, &fl4, 0, oif, NULL, daddr, mark); 64 if (IS_ERR(dst)) 65 return -EHOSTUNREACH; 66 67 saddr->a4 = fl4.saddr; 68 dst_release(dst); 69 return 0; 70 } 71 72 static int xfrm4_get_tos(const struct flowi *fl) 73 { 74 return IPTOS_RT_MASK & fl->u.ip4.flowi4_tos; /* Strip ECN bits */ 75 } 76 77 static int xfrm4_init_path(struct xfrm_dst *path, struct dst_entry *dst, 78 int nfheader_len) 79 { 80 return 0; 81 } 82 83 static int xfrm4_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 84 const struct flowi *fl) 85 { 86 struct rtable *rt = (struct rtable *)xdst->route; 87 const struct flowi4 *fl4 = &fl->u.ip4; 88 89 xdst->u.rt.rt_iif = fl4->flowi4_iif; 90 91 xdst->u.dst.dev = dev; 92 dev_hold(dev); 93 94 /* Sheit... I remember I did this right. Apparently, 95 * it was magically lost, so this code needs audit */ 96 xdst->u.rt.rt_is_input = rt->rt_is_input; 97 xdst->u.rt.rt_flags = rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST | 98 RTCF_LOCAL); 99 xdst->u.rt.rt_type = rt->rt_type; 100 xdst->u.rt.rt_gw_family = rt->rt_gw_family; 101 if (rt->rt_gw_family == AF_INET) 102 xdst->u.rt.rt_gw4 = rt->rt_gw4; 103 else if (rt->rt_gw_family == AF_INET6) 104 xdst->u.rt.rt_gw6 = rt->rt_gw6; 105 xdst->u.rt.rt_pmtu = rt->rt_pmtu; 106 xdst->u.rt.rt_mtu_locked = rt->rt_mtu_locked; 107 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached); 108 rt_add_uncached_list(&xdst->u.rt); 109 110 return 0; 111 } 112 113 static void 114 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse) 115 { 116 const struct iphdr *iph = ip_hdr(skb); 117 u8 *xprth = skb_network_header(skb) + iph->ihl * 4; 118 struct flowi4 *fl4 = &fl->u.ip4; 119 int oif = 0; 120 121 if (skb_dst(skb)) 122 oif = skb_dst(skb)->dev->ifindex; 123 124 memset(fl4, 0, sizeof(struct flowi4)); 125 fl4->flowi4_mark = skb->mark; 126 fl4->flowi4_oif = reverse ? skb->skb_iif : oif; 127 128 if (!ip_is_fragment(iph)) { 129 switch (iph->protocol) { 130 case IPPROTO_UDP: 131 case IPPROTO_UDPLITE: 132 case IPPROTO_TCP: 133 case IPPROTO_SCTP: 134 case IPPROTO_DCCP: 135 if (xprth + 4 < skb->data || 136 pskb_may_pull(skb, xprth + 4 - skb->data)) { 137 __be16 *ports; 138 139 xprth = skb_network_header(skb) + iph->ihl * 4; 140 ports = (__be16 *)xprth; 141 142 fl4->fl4_sport = ports[!!reverse]; 143 fl4->fl4_dport = ports[!reverse]; 144 } 145 break; 146 147 case IPPROTO_ICMP: 148 if (xprth + 2 < skb->data || 149 pskb_may_pull(skb, xprth + 2 - skb->data)) { 150 u8 *icmp; 151 152 xprth = skb_network_header(skb) + iph->ihl * 4; 153 icmp = xprth; 154 155 fl4->fl4_icmp_type = icmp[0]; 156 fl4->fl4_icmp_code = icmp[1]; 157 } 158 break; 159 160 case IPPROTO_ESP: 161 if (xprth + 4 < skb->data || 162 pskb_may_pull(skb, xprth + 4 - skb->data)) { 163 __be32 *ehdr; 164 165 xprth = skb_network_header(skb) + iph->ihl * 4; 166 ehdr = (__be32 *)xprth; 167 168 fl4->fl4_ipsec_spi = ehdr[0]; 169 } 170 break; 171 172 case IPPROTO_AH: 173 if (xprth + 8 < skb->data || 174 pskb_may_pull(skb, xprth + 8 - skb->data)) { 175 __be32 *ah_hdr; 176 177 xprth = skb_network_header(skb) + iph->ihl * 4; 178 ah_hdr = (__be32 *)xprth; 179 180 fl4->fl4_ipsec_spi = ah_hdr[1]; 181 } 182 break; 183 184 case IPPROTO_COMP: 185 if (xprth + 4 < skb->data || 186 pskb_may_pull(skb, xprth + 4 - skb->data)) { 187 __be16 *ipcomp_hdr; 188 189 xprth = skb_network_header(skb) + iph->ihl * 4; 190 ipcomp_hdr = (__be16 *)xprth; 191 192 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); 193 } 194 break; 195 196 case IPPROTO_GRE: 197 if (xprth + 12 < skb->data || 198 pskb_may_pull(skb, xprth + 12 - skb->data)) { 199 __be16 *greflags; 200 __be32 *gre_hdr; 201 202 xprth = skb_network_header(skb) + iph->ihl * 4; 203 greflags = (__be16 *)xprth; 204 gre_hdr = (__be32 *)xprth; 205 206 if (greflags[0] & GRE_KEY) { 207 if (greflags[0] & GRE_CSUM) 208 gre_hdr++; 209 fl4->fl4_gre_key = gre_hdr[1]; 210 } 211 } 212 break; 213 214 default: 215 fl4->fl4_ipsec_spi = 0; 216 break; 217 } 218 } 219 fl4->flowi4_proto = iph->protocol; 220 fl4->daddr = reverse ? iph->saddr : iph->daddr; 221 fl4->saddr = reverse ? iph->daddr : iph->saddr; 222 fl4->flowi4_tos = iph->tos; 223 } 224 225 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk, 226 struct sk_buff *skb, u32 mtu) 227 { 228 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 229 struct dst_entry *path = xdst->route; 230 231 path->ops->update_pmtu(path, sk, skb, mtu); 232 } 233 234 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk, 235 struct sk_buff *skb) 236 { 237 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 238 struct dst_entry *path = xdst->route; 239 240 path->ops->redirect(path, sk, skb); 241 } 242 243 static void xfrm4_dst_destroy(struct dst_entry *dst) 244 { 245 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 246 247 dst_destroy_metrics_generic(dst); 248 if (xdst->u.rt.rt_uncached_list) 249 rt_del_uncached_list(&xdst->u.rt); 250 xfrm_dst_destroy(xdst); 251 } 252 253 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 254 int unregister) 255 { 256 if (!unregister) 257 return; 258 259 xfrm_dst_ifdown(dst, dev); 260 } 261 262 static struct dst_ops xfrm4_dst_ops_template = { 263 .family = AF_INET, 264 .update_pmtu = xfrm4_update_pmtu, 265 .redirect = xfrm4_redirect, 266 .cow_metrics = dst_cow_metrics_generic, 267 .destroy = xfrm4_dst_destroy, 268 .ifdown = xfrm4_dst_ifdown, 269 .local_out = __ip_local_out, 270 .gc_thresh = 32768, 271 }; 272 273 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = { 274 .dst_ops = &xfrm4_dst_ops_template, 275 .dst_lookup = xfrm4_dst_lookup, 276 .get_saddr = xfrm4_get_saddr, 277 .decode_session = _decode_session4, 278 .get_tos = xfrm4_get_tos, 279 .init_path = xfrm4_init_path, 280 .fill_dst = xfrm4_fill_dst, 281 .blackhole_route = ipv4_blackhole_route, 282 }; 283 284 #ifdef CONFIG_SYSCTL 285 static struct ctl_table xfrm4_policy_table[] = { 286 { 287 .procname = "xfrm4_gc_thresh", 288 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh, 289 .maxlen = sizeof(int), 290 .mode = 0644, 291 .proc_handler = proc_dointvec, 292 }, 293 { } 294 }; 295 296 static __net_init int xfrm4_net_sysctl_init(struct net *net) 297 { 298 struct ctl_table *table; 299 struct ctl_table_header *hdr; 300 301 table = xfrm4_policy_table; 302 if (!net_eq(net, &init_net)) { 303 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL); 304 if (!table) 305 goto err_alloc; 306 307 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh; 308 } 309 310 hdr = register_net_sysctl(net, "net/ipv4", table); 311 if (!hdr) 312 goto err_reg; 313 314 net->ipv4.xfrm4_hdr = hdr; 315 return 0; 316 317 err_reg: 318 if (!net_eq(net, &init_net)) 319 kfree(table); 320 err_alloc: 321 return -ENOMEM; 322 } 323 324 static __net_exit void xfrm4_net_sysctl_exit(struct net *net) 325 { 326 struct ctl_table *table; 327 328 if (!net->ipv4.xfrm4_hdr) 329 return; 330 331 table = net->ipv4.xfrm4_hdr->ctl_table_arg; 332 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr); 333 if (!net_eq(net, &init_net)) 334 kfree(table); 335 } 336 #else /* CONFIG_SYSCTL */ 337 static inline int xfrm4_net_sysctl_init(struct net *net) 338 { 339 return 0; 340 } 341 342 static inline void xfrm4_net_sysctl_exit(struct net *net) 343 { 344 } 345 #endif 346 347 static int __net_init xfrm4_net_init(struct net *net) 348 { 349 int ret; 350 351 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template, 352 sizeof(xfrm4_dst_ops_template)); 353 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops); 354 if (ret) 355 return ret; 356 357 ret = xfrm4_net_sysctl_init(net); 358 if (ret) 359 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops); 360 361 return ret; 362 } 363 364 static void __net_exit xfrm4_net_exit(struct net *net) 365 { 366 xfrm4_net_sysctl_exit(net); 367 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops); 368 } 369 370 static struct pernet_operations __net_initdata xfrm4_net_ops = { 371 .init = xfrm4_net_init, 372 .exit = xfrm4_net_exit, 373 }; 374 375 static void __init xfrm4_policy_init(void) 376 { 377 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET); 378 } 379 380 void __init xfrm4_init(void) 381 { 382 xfrm4_state_init(); 383 xfrm4_policy_init(); 384 xfrm4_protocol_init(); 385 register_pernet_subsys(&xfrm4_net_ops); 386 } 387