net/ipv6/seg6_iptunnel.c

   1 /*
   2  *  SR-IPv6 implementation
   3  *
   4  *  Author:
   5  *  David Lebrun <david.lebrun@uclouvain.be>
   6  *
   7  *
   8  *  This program is free software; you can redistribute it and/or
   9  *        modify it under the terms of the GNU General Public License
  10  *        as published by the Free Software Foundation; either version
  11  *        2 of the License, or (at your option) any later version.
  12  */
  13
  14 #include <linux/types.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/net.h>
  17 #include <linux/module.h>
  18 #include <net/ip.h>
  19 #include <net/lwtunnel.h>
  20 #include <net/netevent.h>
  21 #include <net/netns/generic.h>
  22 #include <net/ip6_fib.h>
  23 #include <net/route.h>
  24 #include <net/seg6.h>
  25 #include <linux/seg6.h>
  26 #include <linux/seg6_iptunnel.h>
  27 #include <net/addrconf.h>
  28 #include <net/ip6_route.h>
  29 #include <net/dst_cache.h>
  30 #ifdef CONFIG_IPV6_SEG6_HMAC
  31 #include <net/seg6_hmac.h>
  32 #endif
  33
  34 struct seg6_lwt {
  35         struct dst_cache cache;
  36         struct seg6_iptunnel_encap tuninfo[0];
  37 };
  38
  39 static inline struct seg6_lwt *seg6_lwt_lwtunnel(struct lwtunnel_state *lwt)
  40 {
  41         return (struct seg6_lwt *)lwt->data;
  42 }
  43
  44 static inline struct seg6_iptunnel_encap *
  45 seg6_encap_lwtunnel(struct lwtunnel_state *lwt)
  46 {
  47         return seg6_lwt_lwtunnel(lwt)->tuninfo;
  48 }
  49
  50 static const struct nla_policy seg6_iptunnel_policy[SEG6_IPTUNNEL_MAX + 1] = {
  51         [SEG6_IPTUNNEL_SRH]     = { .type = NLA_BINARY },
  52 };
  53
  54 static int nla_put_srh(struct sk_buff *skb, int attrtype,
  55                        struct seg6_iptunnel_encap *tuninfo)
  56 {
  57         struct seg6_iptunnel_encap *data;
  58         struct nlattr *nla;
  59         int len;
  60
  61         len = SEG6_IPTUN_ENCAP_SIZE(tuninfo);
  62
  63         nla = nla_reserve(skb, attrtype, len);
  64         if (!nla)
  65                 return -EMSGSIZE;
  66
  67         data = nla_data(nla);
  68         memcpy(data, tuninfo, len);
  69
  70         return 0;
  71 }
  72
  73 static void set_tun_src(struct net *net, struct net_device *dev,
  74                         struct in6_addr *daddr, struct in6_addr *saddr)
  75 {
  76         struct seg6_pernet_data *sdata = seg6_pernet(net);
  77         struct in6_addr *tun_src;
  78
  79         rcu_read_lock();
  80
  81         tun_src = rcu_dereference(sdata->tun_src);
  82
  83         if (!ipv6_addr_any(tun_src)) {
  84                 memcpy(saddr, tun_src, sizeof(struct in6_addr));
  85         } else {
  86                 ipv6_dev_get_saddr(net, dev, daddr, IPV6_PREFER_SRC_PUBLIC,
  87                                    saddr);
  88         }
  89
  90         rcu_read_unlock();
  91 }
  92
  93 /* encapsulate an IPv6 packet within an outer IPv6 header with a given SRH */
  94 int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
  95 {
  96         struct net *net = dev_net(skb_dst(skb)->dev);
  97         struct ipv6hdr *hdr, *inner_hdr;
  98         struct ipv6_sr_hdr *isrh;
  99         int hdrlen, tot_len, err;
 100
 101         hdrlen = (osrh->hdrlen + 1) << 3;
 102         tot_len = hdrlen + sizeof(*hdr);
 103
 104         err = skb_cow_head(skb, tot_len);
 105         if (unlikely(err))
 106                 return err;
 107
 108         inner_hdr = ipv6_hdr(skb);
 109
 110         skb_push(skb, tot_len);
 111         skb_reset_network_header(skb);
 112         skb_mac_header_rebuild(skb);
 113         hdr = ipv6_hdr(skb);
 114
 115         /* inherit tc, flowlabel and hlim
 116          * hlim will be decremented in ip6_forward() afterwards and
 117          * decapsulation will overwrite inner hlim with outer hlim
 118          */
 119
 120         if (skb->protocol == htons(ETH_P_IPV6)) {
 121                 ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
 122                              ip6_flowlabel(inner_hdr));
 123                 hdr->hop_limit = inner_hdr->hop_limit;
 124         } else {
 125                 ip6_flow_hdr(hdr, 0, 0);
 126                 hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
 127         }
 128
 129         hdr->nexthdr = NEXTHDR_ROUTING;
 130
 131         isrh = (void *)hdr + sizeof(*hdr);
 132         memcpy(isrh, osrh, hdrlen);
 133
 134         isrh->nexthdr = proto;
 135
 136         hdr->daddr = isrh->segments[isrh->first_segment];
 137         set_tun_src(net, skb->dev, &hdr->daddr, &hdr->saddr);
 138
 139 #ifdef CONFIG_IPV6_SEG6_HMAC
 140         if (sr_has_hmac(isrh)) {
 141                 err = seg6_push_hmac(net, &hdr->saddr, isrh);
 142                 if (unlikely(err))
 143                         return err;
 144         }
 145 #endif
 146
 147         skb_postpush_rcsum(skb, hdr, tot_len);
 148
 149         return 0;
 150 }
 151 EXPORT_SYMBOL_GPL(seg6_do_srh_encap);
 152
 153 /* insert an SRH within an IPv6 packet, just after the IPv6 header */
 154 int seg6_do_srh_inline(struct sk_buff *skb, struct ipv6_sr_hdr *osrh)
 155 {
 156         struct ipv6hdr *hdr, *oldhdr;
 157         struct ipv6_sr_hdr *isrh;
 158         int hdrlen, err;
 159
 160         hdrlen = (osrh->hdrlen + 1) << 3;
 161
 162         err = skb_cow_head(skb, hdrlen);
 163         if (unlikely(err))
 164                 return err;
 165
 166         oldhdr = ipv6_hdr(skb);
 167
 168         skb_pull(skb, sizeof(struct ipv6hdr));
 169         skb_postpull_rcsum(skb, skb_network_header(skb),
 170                            sizeof(struct ipv6hdr));
 171
 172         skb_push(skb, sizeof(struct ipv6hdr) + hdrlen);
 173         skb_reset_network_header(skb);
 174         skb_mac_header_rebuild(skb);
 175
 176         hdr = ipv6_hdr(skb);
 177
 178         memmove(hdr, oldhdr, sizeof(*hdr));
 179
 180         isrh = (void *)hdr + sizeof(*hdr);
 181         memcpy(isrh, osrh, hdrlen);
 182
 183         isrh->nexthdr = hdr->nexthdr;
 184         hdr->nexthdr = NEXTHDR_ROUTING;
 185
 186         isrh->segments[0] = hdr->daddr;
 187         hdr->daddr = isrh->segments[isrh->first_segment];
 188
 189 #ifdef CONFIG_IPV6_SEG6_HMAC
 190         if (sr_has_hmac(isrh)) {
 191                 struct net *net = dev_net(skb_dst(skb)->dev);
 192
 193                 err = seg6_push_hmac(net, &hdr->saddr, isrh);
 194                 if (unlikely(err))
 195                         return err;
 196         }
 197 #endif
 198
 199         skb_postpush_rcsum(skb, hdr, sizeof(struct ipv6hdr) + hdrlen);
 200
 201         return 0;
 202 }
 203 EXPORT_SYMBOL_GPL(seg6_do_srh_inline);
 204
 205 static int seg6_do_srh(struct sk_buff *skb)
 206 {
 207         struct dst_entry *dst = skb_dst(skb);
 208         struct seg6_iptunnel_encap *tinfo;
 209         int proto, err = 0;
 210
 211         tinfo = seg6_encap_lwtunnel(dst->lwtstate);
 212
 213         if (likely(!skb->encapsulation)) {
 214                 skb_reset_inner_headers(skb);
 215                 skb->encapsulation = 1;
 216         }
 217
 218         switch (tinfo->mode) {
 219         case SEG6_IPTUN_MODE_INLINE:
 220                 if (skb->protocol != htons(ETH_P_IPV6))
 221                         return -EINVAL;
 222
 223                 err = seg6_do_srh_inline(skb, tinfo->srh);
 224                 if (err)
 225                         return err;
 226
 227                 skb_reset_inner_headers(skb);
 228                 break;
 229         case SEG6_IPTUN_MODE_ENCAP:
 230                 if (skb->protocol == htons(ETH_P_IPV6))
 231                         proto = IPPROTO_IPV6;
 232                 else if (skb->protocol == htons(ETH_P_IP))
 233                         proto = IPPROTO_IPIP;
 234                 else
 235                         return -EINVAL;
 236
 237                 err = seg6_do_srh_encap(skb, tinfo->srh, proto);
 238                 if (err)
 239                         return err;
 240
 241                 skb->protocol = htons(ETH_P_IPV6);
 242                 break;
 243         case SEG6_IPTUN_MODE_L2ENCAP:
 244                 if (!skb_mac_header_was_set(skb))
 245                         return -EINVAL;
 246
 247                 if (pskb_expand_head(skb, skb->mac_len, 0, GFP_ATOMIC) < 0)
 248                         return -ENOMEM;
 249
 250                 skb_mac_header_rebuild(skb);
 251                 skb_push(skb, skb->mac_len);
 252
 253                 err = seg6_do_srh_encap(skb, tinfo->srh, NEXTHDR_NONE);
 254                 if (err)
 255                         return err;
 256
 257                 skb->protocol = htons(ETH_P_IPV6);
 258                 break;
 259         }
 260
 261         ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
 262         skb_set_transport_header(skb, sizeof(struct ipv6hdr));
 263
 264         skb_set_inner_protocol(skb, skb->protocol);
 265
 266         return 0;
 267 }
 268
 269 static int seg6_input(struct sk_buff *skb)
 270 {
 271         struct dst_entry *orig_dst = skb_dst(skb);
 272         struct dst_entry *dst = NULL;
 273         struct seg6_lwt *slwt;
 274         int err;
 275
 276         err = seg6_do_srh(skb);
 277         if (unlikely(err)) {
 278                 kfree_skb(skb);
 279                 return err;
 280         }
 281
 282         slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
 283
 284         preempt_disable();
 285         dst = dst_cache_get(&slwt->cache);
 286         preempt_enable();
 287
 288         skb_dst_drop(skb);
 289
 290         if (!dst) {
 291                 ip6_route_input(skb);
 292                 dst = skb_dst(skb);
 293                 if (!dst->error) {
 294                         preempt_disable();
 295                         dst_cache_set_ip6(&slwt->cache, dst,
 296                                           &ipv6_hdr(skb)->saddr);
 297                         preempt_enable();
 298                 }
 299         } else {
 300                 skb_dst_set(skb, dst);
 301         }
 302
 303         err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
 304         if (unlikely(err))
 305                 return err;
 306
 307         return dst_input(skb);
 308 }
 309
 310 static int seg6_output(struct net *net, struct sock *sk, struct sk_buff *skb)
 311 {
 312         struct dst_entry *orig_dst = skb_dst(skb);
 313         struct dst_entry *dst = NULL;
 314         struct seg6_lwt *slwt;
 315         int err = -EINVAL;
 316
 317         err = seg6_do_srh(skb);
 318         if (unlikely(err))
 319                 goto drop;
 320
 321         slwt = seg6_lwt_lwtunnel(orig_dst->lwtstate);
 322
 323         preempt_disable();
 324         dst = dst_cache_get(&slwt->cache);
 325         preempt_enable();
 326
 327         if (unlikely(!dst)) {
 328                 struct ipv6hdr *hdr = ipv6_hdr(skb);
 329                 struct flowi6 fl6;
 330
 331                 fl6.daddr = hdr->daddr;
 332                 fl6.saddr = hdr->saddr;
 333                 fl6.flowlabel = ip6_flowinfo(hdr);
 334                 fl6.flowi6_mark = skb->mark;
 335                 fl6.flowi6_proto = hdr->nexthdr;
 336
 337                 dst = ip6_route_output(net, NULL, &fl6);
 338                 if (dst->error) {
 339                         err = dst->error;
 340                         dst_release(dst);
 341                         goto drop;
 342                 }
 343
 344                 preempt_disable();
 345                 dst_cache_set_ip6(&slwt->cache, dst, &fl6.saddr);
 346                 preempt_enable();
 347         }
 348
 349         skb_dst_drop(skb);
 350         skb_dst_set(skb, dst);
 351
 352         err = skb_cow_head(skb, LL_RESERVED_SPACE(dst->dev));
 353         if (unlikely(err))
 354                 goto drop;
 355
 356         return dst_output(net, sk, skb);
 357 drop:
 358         kfree_skb(skb);
 359         return err;
 360 }
 361
 362 static int seg6_build_state(struct nlattr *nla,
 363                             unsigned int family, const void *cfg,
 364                             struct lwtunnel_state **ts,
 365                             struct netlink_ext_ack *extack)
 366 {
 367         struct nlattr *tb[SEG6_IPTUNNEL_MAX + 1];
 368         struct seg6_iptunnel_encap *tuninfo;
 369         struct lwtunnel_state *newts;
 370         int tuninfo_len, min_size;
 371         struct seg6_lwt *slwt;
 372         int err;
 373
 374         if (family != AF_INET && family != AF_INET6)
 375                 return -EINVAL;
 376
 377         err = nla_parse_nested(tb, SEG6_IPTUNNEL_MAX, nla,
 378                                seg6_iptunnel_policy, extack);
 379
 380         if (err < 0)
 381                 return err;
 382
 383         if (!tb[SEG6_IPTUNNEL_SRH])
 384                 return -EINVAL;
 385
 386         tuninfo = nla_data(tb[SEG6_IPTUNNEL_SRH]);
 387         tuninfo_len = nla_len(tb[SEG6_IPTUNNEL_SRH]);
 388
 389         /* tuninfo must contain at least the iptunnel encap structure,
 390          * the SRH and one segment
 391          */
 392         min_size = sizeof(*tuninfo) + sizeof(struct ipv6_sr_hdr) +
 393                    sizeof(struct in6_addr);
 394         if (tuninfo_len < min_size)
 395                 return -EINVAL;
 396
 397         switch (tuninfo->mode) {
 398         case SEG6_IPTUN_MODE_INLINE:
 399                 if (family != AF_INET6)
 400                         return -EINVAL;
 401
 402                 break;
 403         case SEG6_IPTUN_MODE_ENCAP:
 404                 break;
 405         case SEG6_IPTUN_MODE_L2ENCAP:
 406                 break;
 407         default:
 408                 return -EINVAL;
 409         }
 410
 411         /* verify that SRH is consistent */
 412         if (!seg6_validate_srh(tuninfo->srh, tuninfo_len - sizeof(*tuninfo)))
 413                 return -EINVAL;
 414
 415         newts = lwtunnel_state_alloc(tuninfo_len + sizeof(*slwt));
 416         if (!newts)
 417                 return -ENOMEM;
 418
 419         slwt = seg6_lwt_lwtunnel(newts);
 420
 421         err = dst_cache_init(&slwt->cache, GFP_KERNEL);
 422         if (err) {
 423                 kfree(newts);
 424                 return err;
 425         }
 426
 427         memcpy(&slwt->tuninfo, tuninfo, tuninfo_len);
 428
 429         newts->type = LWTUNNEL_ENCAP_SEG6;
 430         newts->flags |= LWTUNNEL_STATE_INPUT_REDIRECT;
 431
 432         if (tuninfo->mode != SEG6_IPTUN_MODE_L2ENCAP)
 433                 newts->flags |= LWTUNNEL_STATE_OUTPUT_REDIRECT;
 434
 435         newts->headroom = seg6_lwt_headroom(tuninfo);
 436
 437         *ts = newts;
 438
 439         return 0;
 440 }
 441
 442 static void seg6_destroy_state(struct lwtunnel_state *lwt)
 443 {
 444         dst_cache_destroy(&seg6_lwt_lwtunnel(lwt)->cache);
 445 }
 446
 447 static int seg6_fill_encap_info(struct sk_buff *skb,
 448                                 struct lwtunnel_state *lwtstate)
 449 {
 450         struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
 451
 452         if (nla_put_srh(skb, SEG6_IPTUNNEL_SRH, tuninfo))
 453                 return -EMSGSIZE;
 454
 455         return 0;
 456 }
 457
 458 static int seg6_encap_nlsize(struct lwtunnel_state *lwtstate)
 459 {
 460         struct seg6_iptunnel_encap *tuninfo = seg6_encap_lwtunnel(lwtstate);
 461
 462         return nla_total_size(SEG6_IPTUN_ENCAP_SIZE(tuninfo));
 463 }
 464
 465 static int seg6_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
 466 {
 467         struct seg6_iptunnel_encap *a_hdr = seg6_encap_lwtunnel(a);
 468         struct seg6_iptunnel_encap *b_hdr = seg6_encap_lwtunnel(b);
 469         int len = SEG6_IPTUN_ENCAP_SIZE(a_hdr);
 470
 471         if (len != SEG6_IPTUN_ENCAP_SIZE(b_hdr))
 472                 return 1;
 473
 474         return memcmp(a_hdr, b_hdr, len);
 475 }
 476
 477 static const struct lwtunnel_encap_ops seg6_iptun_ops = {
 478         .build_state = seg6_build_state,
 479         .destroy_state = seg6_destroy_state,
 480         .output = seg6_output,
 481         .input = seg6_input,
 482         .fill_encap = seg6_fill_encap_info,
 483         .get_encap_size = seg6_encap_nlsize,
 484         .cmp_encap = seg6_encap_cmp,
 485         .owner = THIS_MODULE,
 486 };
 487
 488 int __init seg6_iptunnel_init(void)
 489 {
 490         return lwtunnel_encap_add_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
 491 }
 492
 493 void seg6_iptunnel_exit(void)
 494 {
 495         lwtunnel_encap_del_ops(&seg6_iptun_ops, LWTUNNEL_ENCAP_SEG6);
 496 }