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_gateway = rt->rt_gateway; 101 xdst->u.rt.rt_uses_gateway = rt->rt_uses_gateway; 102 xdst->u.rt.rt_pmtu = rt->rt_pmtu; 103 xdst->u.rt.rt_table_id = rt->rt_table_id; 104 INIT_LIST_HEAD(&xdst->u.rt.rt_uncached); 105 106 return 0; 107 } 108 109 static void 110 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse) 111 { 112 const struct iphdr *iph = ip_hdr(skb); 113 u8 *xprth = skb_network_header(skb) + iph->ihl * 4; 114 struct flowi4 *fl4 = &fl->u.ip4; 115 int oif = 0; 116 117 if (skb_dst(skb)) 118 oif = skb_dst(skb)->dev->ifindex; 119 120 memset(fl4, 0, sizeof(struct flowi4)); 121 fl4->flowi4_mark = skb->mark; 122 fl4->flowi4_oif = reverse ? skb->skb_iif : oif; 123 124 if (!ip_is_fragment(iph)) { 125 switch (iph->protocol) { 126 case IPPROTO_UDP: 127 case IPPROTO_UDPLITE: 128 case IPPROTO_TCP: 129 case IPPROTO_SCTP: 130 case IPPROTO_DCCP: 131 if (xprth + 4 < skb->data || 132 pskb_may_pull(skb, xprth + 4 - skb->data)) { 133 __be16 *ports; 134 135 xprth = skb_network_header(skb) + iph->ihl * 4; 136 ports = (__be16 *)xprth; 137 138 fl4->fl4_sport = ports[!!reverse]; 139 fl4->fl4_dport = ports[!reverse]; 140 } 141 break; 142 143 case IPPROTO_ICMP: 144 if (xprth + 2 < skb->data || 145 pskb_may_pull(skb, xprth + 2 - skb->data)) { 146 u8 *icmp; 147 148 xprth = skb_network_header(skb) + iph->ihl * 4; 149 icmp = xprth; 150 151 fl4->fl4_icmp_type = icmp[0]; 152 fl4->fl4_icmp_code = icmp[1]; 153 } 154 break; 155 156 case IPPROTO_ESP: 157 if (xprth + 4 < skb->data || 158 pskb_may_pull(skb, xprth + 4 - skb->data)) { 159 __be32 *ehdr; 160 161 xprth = skb_network_header(skb) + iph->ihl * 4; 162 ehdr = (__be32 *)xprth; 163 164 fl4->fl4_ipsec_spi = ehdr[0]; 165 } 166 break; 167 168 case IPPROTO_AH: 169 if (xprth + 8 < skb->data || 170 pskb_may_pull(skb, xprth + 8 - skb->data)) { 171 __be32 *ah_hdr; 172 173 xprth = skb_network_header(skb) + iph->ihl * 4; 174 ah_hdr = (__be32 *)xprth; 175 176 fl4->fl4_ipsec_spi = ah_hdr[1]; 177 } 178 break; 179 180 case IPPROTO_COMP: 181 if (xprth + 4 < skb->data || 182 pskb_may_pull(skb, xprth + 4 - skb->data)) { 183 __be16 *ipcomp_hdr; 184 185 xprth = skb_network_header(skb) + iph->ihl * 4; 186 ipcomp_hdr = (__be16 *)xprth; 187 188 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1])); 189 } 190 break; 191 192 case IPPROTO_GRE: 193 if (xprth + 12 < skb->data || 194 pskb_may_pull(skb, xprth + 12 - skb->data)) { 195 __be16 *greflags; 196 __be32 *gre_hdr; 197 198 xprth = skb_network_header(skb) + iph->ihl * 4; 199 greflags = (__be16 *)xprth; 200 gre_hdr = (__be32 *)xprth; 201 202 if (greflags[0] & GRE_KEY) { 203 if (greflags[0] & GRE_CSUM) 204 gre_hdr++; 205 fl4->fl4_gre_key = gre_hdr[1]; 206 } 207 } 208 break; 209 210 default: 211 fl4->fl4_ipsec_spi = 0; 212 break; 213 } 214 } 215 fl4->flowi4_proto = iph->protocol; 216 fl4->daddr = reverse ? iph->saddr : iph->daddr; 217 fl4->saddr = reverse ? iph->daddr : iph->saddr; 218 fl4->flowi4_tos = iph->tos; 219 } 220 221 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk, 222 struct sk_buff *skb, u32 mtu) 223 { 224 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 225 struct dst_entry *path = xdst->route; 226 227 path->ops->update_pmtu(path, sk, skb, mtu); 228 } 229 230 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk, 231 struct sk_buff *skb) 232 { 233 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 234 struct dst_entry *path = xdst->route; 235 236 path->ops->redirect(path, sk, skb); 237 } 238 239 static void xfrm4_dst_destroy(struct dst_entry *dst) 240 { 241 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 242 243 dst_destroy_metrics_generic(dst); 244 245 xfrm_dst_destroy(xdst); 246 } 247 248 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 249 int unregister) 250 { 251 if (!unregister) 252 return; 253 254 xfrm_dst_ifdown(dst, dev); 255 } 256 257 static struct dst_ops xfrm4_dst_ops_template = { 258 .family = AF_INET, 259 .update_pmtu = xfrm4_update_pmtu, 260 .redirect = xfrm4_redirect, 261 .cow_metrics = dst_cow_metrics_generic, 262 .destroy = xfrm4_dst_destroy, 263 .ifdown = xfrm4_dst_ifdown, 264 .local_out = __ip_local_out, 265 .gc_thresh = 32768, 266 }; 267 268 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = { 269 .dst_ops = &xfrm4_dst_ops_template, 270 .dst_lookup = xfrm4_dst_lookup, 271 .get_saddr = xfrm4_get_saddr, 272 .decode_session = _decode_session4, 273 .get_tos = xfrm4_get_tos, 274 .init_path = xfrm4_init_path, 275 .fill_dst = xfrm4_fill_dst, 276 .blackhole_route = ipv4_blackhole_route, 277 }; 278 279 #ifdef CONFIG_SYSCTL 280 static struct ctl_table xfrm4_policy_table[] = { 281 { 282 .procname = "xfrm4_gc_thresh", 283 .data = &init_net.xfrm.xfrm4_dst_ops.gc_thresh, 284 .maxlen = sizeof(int), 285 .mode = 0644, 286 .proc_handler = proc_dointvec, 287 }, 288 { } 289 }; 290 291 static __net_init int xfrm4_net_sysctl_init(struct net *net) 292 { 293 struct ctl_table *table; 294 struct ctl_table_header *hdr; 295 296 table = xfrm4_policy_table; 297 if (!net_eq(net, &init_net)) { 298 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL); 299 if (!table) 300 goto err_alloc; 301 302 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh; 303 } 304 305 hdr = register_net_sysctl(net, "net/ipv4", table); 306 if (!hdr) 307 goto err_reg; 308 309 net->ipv4.xfrm4_hdr = hdr; 310 return 0; 311 312 err_reg: 313 if (!net_eq(net, &init_net)) 314 kfree(table); 315 err_alloc: 316 return -ENOMEM; 317 } 318 319 static __net_exit void xfrm4_net_sysctl_exit(struct net *net) 320 { 321 struct ctl_table *table; 322 323 if (!net->ipv4.xfrm4_hdr) 324 return; 325 326 table = net->ipv4.xfrm4_hdr->ctl_table_arg; 327 unregister_net_sysctl_table(net->ipv4.xfrm4_hdr); 328 if (!net_eq(net, &init_net)) 329 kfree(table); 330 } 331 #else /* CONFIG_SYSCTL */ 332 static inline int xfrm4_net_sysctl_init(struct net *net) 333 { 334 return 0; 335 } 336 337 static inline void xfrm4_net_sysctl_exit(struct net *net) 338 { 339 } 340 #endif 341 342 static int __net_init xfrm4_net_init(struct net *net) 343 { 344 int ret; 345 346 memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template, 347 sizeof(xfrm4_dst_ops_template)); 348 ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops); 349 if (ret) 350 return ret; 351 352 ret = xfrm4_net_sysctl_init(net); 353 if (ret) 354 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops); 355 356 return ret; 357 } 358 359 static void __net_exit xfrm4_net_exit(struct net *net) 360 { 361 xfrm4_net_sysctl_exit(net); 362 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops); 363 } 364 365 static struct pernet_operations __net_initdata xfrm4_net_ops = { 366 .init = xfrm4_net_init, 367 .exit = xfrm4_net_exit, 368 }; 369 370 static void __init xfrm4_policy_init(void) 371 { 372 xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET); 373 } 374 375 void __init xfrm4_init(void) 376 { 377 xfrm4_state_init(); 378 xfrm4_policy_init(); 379 xfrm4_protocol_init(); 380 register_pernet_subsys(&xfrm4_net_ops); 381 } 382 383