net/ipv4/xfrm4_policy.c

   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_mtu_locked = rt->rt_mtu_locked;
 104         xdst->u.rt.rt_table_id = rt->rt_table_id;
 105         INIT_LIST_HEAD(&xdst->u.rt.rt_uncached);
 106         rt_add_uncached_list(&xdst->u.rt);
 107
 108         return 0;
 109 }
 110
 111 static void
 112 _decode_session4(struct sk_buff *skb, struct flowi *fl, int reverse)
 113 {
 114         const struct iphdr *iph = ip_hdr(skb);
 115         u8 *xprth = skb_network_header(skb) + iph->ihl * 4;
 116         struct flowi4 *fl4 = &fl->u.ip4;
 117         int oif = 0;
 118
 119         if (skb_dst(skb))
 120                 oif = skb_dst(skb)->dev->ifindex;
 121
 122         memset(fl4, 0, sizeof(struct flowi4));
 123         fl4->flowi4_mark = skb->mark;
 124         fl4->flowi4_oif = reverse ? skb->skb_iif : oif;
 125
 126         if (!ip_is_fragment(iph)) {
 127                 switch (iph->protocol) {
 128                 case IPPROTO_UDP:
 129                 case IPPROTO_UDPLITE:
 130                 case IPPROTO_TCP:
 131                 case IPPROTO_SCTP:
 132                 case IPPROTO_DCCP:
 133                         if (xprth + 4 < skb->data ||
 134                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 135                                 __be16 *ports;
 136
 137                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 138                                 ports = (__be16 *)xprth;
 139
 140                                 fl4->fl4_sport = ports[!!reverse];
 141                                 fl4->fl4_dport = ports[!reverse];
 142                         }
 143                         break;
 144
 145                 case IPPROTO_ICMP:
 146                         if (xprth + 2 < skb->data ||
 147                             pskb_may_pull(skb, xprth + 2 - skb->data)) {
 148                                 u8 *icmp;
 149
 150                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 151                                 icmp = xprth;
 152
 153                                 fl4->fl4_icmp_type = icmp[0];
 154                                 fl4->fl4_icmp_code = icmp[1];
 155                         }
 156                         break;
 157
 158                 case IPPROTO_ESP:
 159                         if (xprth + 4 < skb->data ||
 160                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 161                                 __be32 *ehdr;
 162
 163                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 164                                 ehdr = (__be32 *)xprth;
 165
 166                                 fl4->fl4_ipsec_spi = ehdr[0];
 167                         }
 168                         break;
 169
 170                 case IPPROTO_AH:
 171                         if (xprth + 8 < skb->data ||
 172                             pskb_may_pull(skb, xprth + 8 - skb->data)) {
 173                                 __be32 *ah_hdr;
 174
 175                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 176                                 ah_hdr = (__be32 *)xprth;
 177
 178                                 fl4->fl4_ipsec_spi = ah_hdr[1];
 179                         }
 180                         break;
 181
 182                 case IPPROTO_COMP:
 183                         if (xprth + 4 < skb->data ||
 184                             pskb_may_pull(skb, xprth + 4 - skb->data)) {
 185                                 __be16 *ipcomp_hdr;
 186
 187                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 188                                 ipcomp_hdr = (__be16 *)xprth;
 189
 190                                 fl4->fl4_ipsec_spi = htonl(ntohs(ipcomp_hdr[1]));
 191                         }
 192                         break;
 193
 194                 case IPPROTO_GRE:
 195                         if (xprth + 12 < skb->data ||
 196                             pskb_may_pull(skb, xprth + 12 - skb->data)) {
 197                                 __be16 *greflags;
 198                                 __be32 *gre_hdr;
 199
 200                                 xprth = skb_network_header(skb) + iph->ihl * 4;
 201                                 greflags = (__be16 *)xprth;
 202                                 gre_hdr = (__be32 *)xprth;
 203
 204                                 if (greflags[0] & GRE_KEY) {
 205                                         if (greflags[0] & GRE_CSUM)
 206                                                 gre_hdr++;
 207                                         fl4->fl4_gre_key = gre_hdr[1];
 208                                 }
 209                         }
 210                         break;
 211
 212                 default:
 213                         fl4->fl4_ipsec_spi = 0;
 214                         break;
 215                 }
 216         }
 217         fl4->flowi4_proto = iph->protocol;
 218         fl4->daddr = reverse ? iph->saddr : iph->daddr;
 219         fl4->saddr = reverse ? iph->daddr : iph->saddr;
 220         fl4->flowi4_tos = iph->tos;
 221 }
 222
 223 static void xfrm4_update_pmtu(struct dst_entry *dst, struct sock *sk,
 224                               struct sk_buff *skb, u32 mtu)
 225 {
 226         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 227         struct dst_entry *path = xdst->route;
 228
 229         path->ops->update_pmtu(path, sk, skb, mtu);
 230 }
 231
 232 static void xfrm4_redirect(struct dst_entry *dst, struct sock *sk,
 233                            struct sk_buff *skb)
 234 {
 235         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 236         struct dst_entry *path = xdst->route;
 237
 238         path->ops->redirect(path, sk, skb);
 239 }
 240
 241 static void xfrm4_dst_destroy(struct dst_entry *dst)
 242 {
 243         struct xfrm_dst *xdst = (struct xfrm_dst *)dst;
 244
 245         dst_destroy_metrics_generic(dst);
 246         if (xdst->u.rt.rt_uncached_list)
 247                 rt_del_uncached_list(&xdst->u.rt);
 248         xfrm_dst_destroy(xdst);
 249 }
 250
 251 static void xfrm4_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 252                              int unregister)
 253 {
 254         if (!unregister)
 255                 return;
 256
 257         xfrm_dst_ifdown(dst, dev);
 258 }
 259
 260 static struct dst_ops xfrm4_dst_ops_template = {
 261         .family =               AF_INET,
 262         .update_pmtu =          xfrm4_update_pmtu,
 263         .redirect =             xfrm4_redirect,
 264         .cow_metrics =          dst_cow_metrics_generic,
 265         .destroy =              xfrm4_dst_destroy,
 266         .ifdown =               xfrm4_dst_ifdown,
 267         .local_out =            __ip_local_out,
 268         .gc_thresh =            32768,
 269 };
 270
 271 static const struct xfrm_policy_afinfo xfrm4_policy_afinfo = {
 272         .dst_ops =              &xfrm4_dst_ops_template,
 273         .dst_lookup =           xfrm4_dst_lookup,
 274         .get_saddr =            xfrm4_get_saddr,
 275         .decode_session =       _decode_session4,
 276         .get_tos =              xfrm4_get_tos,
 277         .init_path =            xfrm4_init_path,
 278         .fill_dst =             xfrm4_fill_dst,
 279         .blackhole_route =      ipv4_blackhole_route,
 280 };
 281
 282 #ifdef CONFIG_SYSCTL
 283 static struct ctl_table xfrm4_policy_table[] = {
 284         {
 285                 .procname       = "xfrm4_gc_thresh",
 286                 .data           = &init_net.xfrm.xfrm4_dst_ops.gc_thresh,
 287                 .maxlen         = sizeof(int),
 288                 .mode           = 0644,
 289                 .proc_handler   = proc_dointvec,
 290         },
 291         { }
 292 };
 293
 294 static __net_init int xfrm4_net_sysctl_init(struct net *net)
 295 {
 296         struct ctl_table *table;
 297         struct ctl_table_header *hdr;
 298
 299         table = xfrm4_policy_table;
 300         if (!net_eq(net, &init_net)) {
 301                 table = kmemdup(table, sizeof(xfrm4_policy_table), GFP_KERNEL);
 302                 if (!table)
 303                         goto err_alloc;
 304
 305                 table[0].data = &net->xfrm.xfrm4_dst_ops.gc_thresh;
 306         }
 307
 308         hdr = register_net_sysctl(net, "net/ipv4", table);
 309         if (!hdr)
 310                 goto err_reg;
 311
 312         net->ipv4.xfrm4_hdr = hdr;
 313         return 0;
 314
 315 err_reg:
 316         if (!net_eq(net, &init_net))
 317                 kfree(table);
 318 err_alloc:
 319         return -ENOMEM;
 320 }
 321
 322 static __net_exit void xfrm4_net_sysctl_exit(struct net *net)
 323 {
 324         struct ctl_table *table;
 325
 326         if (!net->ipv4.xfrm4_hdr)
 327                 return;
 328
 329         table = net->ipv4.xfrm4_hdr->ctl_table_arg;
 330         unregister_net_sysctl_table(net->ipv4.xfrm4_hdr);
 331         if (!net_eq(net, &init_net))
 332                 kfree(table);
 333 }
 334 #else /* CONFIG_SYSCTL */
 335 static inline int xfrm4_net_sysctl_init(struct net *net)
 336 {
 337         return 0;
 338 }
 339
 340 static inline void xfrm4_net_sysctl_exit(struct net *net)
 341 {
 342 }
 343 #endif
 344
 345 static int __net_init xfrm4_net_init(struct net *net)
 346 {
 347         int ret;
 348
 349         memcpy(&net->xfrm.xfrm4_dst_ops, &xfrm4_dst_ops_template,
 350                sizeof(xfrm4_dst_ops_template));
 351         ret = dst_entries_init(&net->xfrm.xfrm4_dst_ops);
 352         if (ret)
 353                 return ret;
 354
 355         ret = xfrm4_net_sysctl_init(net);
 356         if (ret)
 357                 dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 358
 359         return ret;
 360 }
 361
 362 static void __net_exit xfrm4_net_exit(struct net *net)
 363 {
 364         xfrm4_net_sysctl_exit(net);
 365         dst_entries_destroy(&net->xfrm.xfrm4_dst_ops);
 366 }
 367
 368 static struct pernet_operations __net_initdata xfrm4_net_ops = {
 369         .init   = xfrm4_net_init,
 370         .exit   = xfrm4_net_exit,
 371 };
 372
 373 static void __init xfrm4_policy_init(void)
 374 {
 375         xfrm_policy_register_afinfo(&xfrm4_policy_afinfo, AF_INET);
 376 }
 377
 378 void __init xfrm4_init(void)
 379 {
 380         xfrm4_state_init();
 381         xfrm4_policy_init();
 382         xfrm4_protocol_init();
 383         register_pernet_subsys(&xfrm4_net_ops);
 384 }
 385