datapath/vport-lisp.c

   1 /*
   2  * Copyright (c) 2011 Nicira, Inc.
   3  * Copyright (c) 2013 Cisco Systems, Inc.
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of version 2 of the GNU General Public
   7  * License as published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it will be useful, but
  10  * WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  12  * General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License
  15  * along with this program; if not, write to the Free Software
  16  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  17  * 02110-1301, USA
  18  */
  19
  20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  21
  22 #include <linux/version.h>
  23
  24 #include <linux/in.h>
  25 #include <linux/ip.h>
  26 #include <linux/net.h>
  27 #include <linux/rculist.h>
  28 #include <linux/udp.h>
  29
  30 #include <net/icmp.h>
  31 #include <net/ip.h>
  32 #include <net/route.h>
  33 #include <net/udp.h>
  34 #include <net/udp_tunnel.h>
  35 #include <net/xfrm.h>
  36
  37 #include "datapath.h"
  38 #include "gso.h"
  39 #include "vport.h"
  40
  41 /*
  42  *  LISP encapsulation header:
  43  *
  44  *  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  45  *  |N|L|E|V|I|flags|            Nonce/Map-Version                  |
  46  *  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  47  *  |                 Instance ID/Locator Status Bits               |
  48  *  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  49  *
  50  */
  51
  52 /**
  53  * struct lisphdr - LISP header
  54  * @nonce_present: Flag indicating the presence of a 24 bit nonce value.
  55  * @locator_status_bits_present: Flag indicating the presence of Locator Status
  56  *                               Bits (LSB).
  57  * @solicit_echo_nonce: Flag indicating the use of the echo noncing mechanism.
  58  * @map_version_present: Flag indicating the use of mapping versioning.
  59  * @instance_id_present: Flag indicating the presence of a 24 bit Instance ID.
  60  * @reserved_flags: 3 bits reserved for future flags.
  61  * @nonce: 24 bit nonce value.
  62  * @map_version: 24 bit mapping version.
  63  * @locator_status_bits: Locator Status Bits: 32 bits when instance_id_present
  64  *                       is not set, 8 bits when it is.
  65  * @instance_id: 24 bit Instance ID
  66  */
  67 struct lisphdr {
  68 #ifdef __LITTLE_ENDIAN_BITFIELD
  69         __u8 reserved_flags:3;
  70         __u8 instance_id_present:1;
  71         __u8 map_version_present:1;
  72         __u8 solicit_echo_nonce:1;
  73         __u8 locator_status_bits_present:1;
  74         __u8 nonce_present:1;
  75 #else
  76         __u8 nonce_present:1;
  77         __u8 locator_status_bits_present:1;
  78         __u8 solicit_echo_nonce:1;
  79         __u8 map_version_present:1;
  80         __u8 instance_id_present:1;
  81         __u8 reserved_flags:3;
  82 #endif
  83         union {
  84                 __u8 nonce[3];
  85                 __u8 map_version[3];
  86         } u1;
  87         union {
  88                 __be32 locator_status_bits;
  89                 struct {
  90                         __u8 instance_id[3];
  91                         __u8 locator_status_bits;
  92                 } word2;
  93         } u2;
  94 };
  95
  96 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
  97
  98 /**
  99  * struct lisp_port - Keeps track of open UDP ports
 100  * @dst_port: lisp UDP port no.
 101  * @list: list element in @lisp_ports.
 102  * @lisp_rcv_socket: The socket created for this port number.
 103  * @name: vport name.
 104  */
 105 struct lisp_port {
 106         __be16 dst_port;
 107         struct list_head list;
 108         struct socket *lisp_rcv_socket;
 109         char name[IFNAMSIZ];
 110 };
 111
 112 static LIST_HEAD(lisp_ports);
 113
 114 static inline struct lisp_port *lisp_vport(const struct vport *vport)
 115 {
 116         return vport_priv(vport);
 117 }
 118
 119 static struct lisp_port *lisp_find_port(struct net *net, __be16 port)
 120 {
 121         struct lisp_port *lisp_port;
 122
 123         list_for_each_entry_rcu(lisp_port, &lisp_ports, list) {
 124                 if (lisp_port->dst_port == port &&
 125                         net_eq(sock_net(lisp_port->lisp_rcv_socket->sk), net))
 126                         return lisp_port;
 127         }
 128
 129         return NULL;
 130 }
 131
 132 static inline struct lisphdr *lisp_hdr(const struct sk_buff *skb)
 133 {
 134         return (struct lisphdr *)(udp_hdr(skb) + 1);
 135 }
 136
 137 /* Convert 64 bit tunnel ID to 24 bit Instance ID. */
 138 static void tunnel_id_to_instance_id(__be64 tun_id, __u8 *iid)
 139 {
 140
 141 #ifdef __BIG_ENDIAN
 142         iid[0] = (__force __u8)(tun_id >> 16);
 143         iid[1] = (__force __u8)(tun_id >> 8);
 144         iid[2] = (__force __u8)tun_id;
 145 #else
 146         iid[0] = (__force __u8)((__force u64)tun_id >> 40);
 147         iid[1] = (__force __u8)((__force u64)tun_id >> 48);
 148         iid[2] = (__force __u8)((__force u64)tun_id >> 56);
 149 #endif
 150 }
 151
 152 /* Convert 24 bit Instance ID to 64 bit tunnel ID. */
 153 static __be64 instance_id_to_tunnel_id(__u8 *iid)
 154 {
 155 #ifdef __BIG_ENDIAN
 156         return (iid[0] << 16) | (iid[1] << 8) | iid[2];
 157 #else
 158         return (__force __be64)(((__force u64)iid[0] << 40) |
 159                                 ((__force u64)iid[1] << 48) |
 160                                 ((__force u64)iid[2] << 56));
 161 #endif
 162 }
 163
 164 /* Compute source UDP port for outgoing packet.
 165  * Currently we use the flow hash.
 166  */
 167 static u16 get_src_port(struct net *net, struct sk_buff *skb)
 168 {
 169         u32 hash = skb_get_hash(skb);
 170         unsigned int range;
 171         int high;
 172         int low;
 173
 174         if (!hash) {
 175                 if (skb->protocol == htons(ETH_P_IP)) {
 176                         struct iphdr *iph;
 177                         int size = (sizeof(iph->saddr) * 2) / sizeof(u32);
 178
 179                         iph = (struct iphdr *) skb_inner_network_header(skb);
 180                         hash = jhash2((const u32 *)&iph->saddr, size, 0);
 181                 } else if (skb->protocol == htons(ETH_P_IPV6)) {
 182                         struct ipv6hdr *ipv6hdr;
 183
 184                         ipv6hdr = (struct ipv6hdr *) skb_inner_network_header(skb);
 185                         hash = jhash2((const u32 *)&ipv6hdr->saddr,
 186                                       (sizeof(struct in6_addr) * 2) / sizeof(u32), 0);
 187                 } else {
 188                         pr_warn_once("LISP inner protocol is not IP when "
 189                                      "calculating hash.\n");
 190                 }
 191         }
 192
 193         inet_get_local_port_range(net, &low, &high);
 194         range = (high - low) + 1;
 195         return (((u64) hash * range) >> 32) + low;
 196 }
 197
 198 static void lisp_build_header(const struct vport *vport,
 199                               struct sk_buff *skb)
 200 {
 201         struct net *net = ovs_dp_get_net(vport->dp);
 202         struct lisp_port *lisp_port = lisp_vport(vport);
 203         struct udphdr *udph = udp_hdr(skb);
 204         struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
 205         const struct ovs_key_ipv4_tunnel *tun_key;
 206
 207         tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
 208         udph->dest = lisp_port->dst_port;
 209         udph->source = htons(get_src_port(net, skb));
 210         udph->check = 0;
 211         udph->len = htons(skb->len - skb_transport_offset(skb));
 212
 213         lisph->nonce_present = 0;       /* We don't support echo nonce algorithm */
 214         lisph->locator_status_bits_present = 1; /* Set LSB */
 215         lisph->solicit_echo_nonce = 0;  /* No echo noncing */
 216         lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
 217         lisph->instance_id_present = 1; /* Store the tun_id as Instance ID  */
 218         lisph->reserved_flags = 0;      /* Reserved flags, set to 0  */
 219
 220         lisph->u1.nonce[0] = 0;
 221         lisph->u1.nonce[1] = 0;
 222         lisph->u1.nonce[2] = 0;
 223
 224         tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
 225         lisph->u2.word2.locator_status_bits = 1;
 226 }
 227
 228 /* Called with rcu_read_lock and BH disabled. */
 229 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 230 {
 231         struct lisp_port *lisp_port;
 232         struct lisphdr *lisph;
 233         struct iphdr *iph, *inner_iph;
 234         struct ovs_tunnel_info tun_info;
 235         __be64 key;
 236         struct ethhdr *ethh;
 237         __be16 protocol;
 238
 239         lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
 240         if (unlikely(!lisp_port))
 241                 goto error;
 242
 243         if (iptunnel_pull_header(skb, LISP_HLEN, 0))
 244                 goto error;
 245
 246         lisph = lisp_hdr(skb);
 247
 248         if (lisph->instance_id_present != 1)
 249                 key = 0;
 250         else
 251                 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
 252
 253         /* Save outer tunnel values */
 254         iph = ip_hdr(skb);
 255         ovs_flow_tun_info_init(&tun_info, iph,
 256                                udp_hdr(skb)->source, udp_hdr(skb)->dest,
 257                                key, TUNNEL_KEY, NULL, 0);
 258
 259         /* Drop non-IP inner packets */
 260         inner_iph = (struct iphdr *)(lisph + 1);
 261         switch (inner_iph->version) {
 262         case 4:
 263                 protocol = htons(ETH_P_IP);
 264                 break;
 265         case 6:
 266                 protocol = htons(ETH_P_IPV6);
 267                 break;
 268         default:
 269                 goto error;
 270         }
 271         skb->protocol = protocol;
 272
 273         /* Add Ethernet header */
 274         ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
 275         memset(ethh, 0, ETH_HLEN);
 276         ethh->h_dest[0] = 0x02;
 277         ethh->h_source[0] = 0x02;
 278         ethh->h_proto = protocol;
 279
 280         ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
 281
 282         ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_info);
 283         goto out;
 284
 285 error:
 286         kfree_skb(skb);
 287 out:
 288         return 0;
 289 }
 290
 291 /* Arbitrary value.  Irrelevant as long as it's not 0 since we set the handler. */
 292 #define UDP_ENCAP_LISP 1
 293 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
 294 {
 295         struct udp_port_cfg udp_conf;
 296         int err;
 297
 298         memset(&udp_conf, 0, sizeof(udp_conf));
 299
 300         udp_conf.family = AF_INET;
 301         udp_conf.local_ip.s_addr = htonl(INADDR_ANY);
 302         udp_conf.local_udp_port = lisp_port->dst_port;
 303
 304         err = udp_sock_create(net, &udp_conf, &lisp_port->lisp_rcv_socket);
 305         if (err < 0) {
 306                 pr_warn("cannot register lisp protocol handler: %d\n", err);
 307                 return err;
 308         }
 309
 310         udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_type = UDP_ENCAP_LISP;
 311         udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_rcv = lisp_rcv;
 312
 313         udp_encap_enable();
 314
 315         return 0;
 316 }
 317
 318 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
 319 {
 320         struct lisp_port *lisp_port = lisp_vport(vport);
 321
 322         if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
 323                 return -EMSGSIZE;
 324         return 0;
 325 }
 326
 327 static void lisp_tnl_destroy(struct vport *vport)
 328 {
 329         struct lisp_port *lisp_port = lisp_vport(vport);
 330
 331         list_del_rcu(&lisp_port->list);
 332         /* Release socket */
 333         sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
 334
 335         ovs_vport_deferred_free(vport);
 336 }
 337
 338 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
 339 {
 340         struct net *net = ovs_dp_get_net(parms->dp);
 341         struct nlattr *options = parms->options;
 342         struct lisp_port *lisp_port;
 343         struct vport *vport;
 344         struct nlattr *a;
 345         int err;
 346         u16 dst_port;
 347
 348         if (!options) {
 349                 err = -EINVAL;
 350                 goto error;
 351         }
 352
 353         a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
 354         if (a && nla_len(a) == sizeof(u16)) {
 355                 dst_port = nla_get_u16(a);
 356         } else {
 357                 /* Require destination port from userspace. */
 358                 err = -EINVAL;
 359                 goto error;
 360         }
 361
 362         /* Verify if we already have a socket created for this port */
 363         if (lisp_find_port(net, htons(dst_port))) {
 364                 err = -EEXIST;
 365                 goto error;
 366         }
 367
 368         vport = ovs_vport_alloc(sizeof(struct lisp_port),
 369                                 &ovs_lisp_vport_ops, parms);
 370         if (IS_ERR(vport))
 371                 return vport;
 372
 373         lisp_port = lisp_vport(vport);
 374         lisp_port->dst_port = htons(dst_port);
 375         strncpy(lisp_port->name, parms->name, IFNAMSIZ);
 376
 377         err = lisp_socket_init(lisp_port, net);
 378         if (err)
 379                 goto error_free;
 380
 381         list_add_tail_rcu(&lisp_port->list, &lisp_ports);
 382         return vport;
 383
 384 error_free:
 385         ovs_vport_free(vport);
 386 error:
 387         return ERR_PTR(err);
 388 }
 389
 390 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,12,0)
 391
 392 static void lisp_fix_segment(struct sk_buff *skb)
 393 {
 394         struct udphdr *udph = udp_hdr(skb);
 395
 396         udph->len = htons(skb->len - skb_transport_offset(skb));
 397 }
 398
 399 static struct sk_buff *handle_offloads(struct sk_buff *skb)
 400 {
 401         return ovs_iptunnel_handle_offloads(skb, false, lisp_fix_segment);
 402 }
 403 #else
 404 static struct sk_buff *handle_offloads(struct sk_buff *skb)
 405 {
 406         int err = 0;
 407
 408         if (skb_is_gso(skb)) {
 409
 410                 if (skb_is_encapsulated(skb)) {
 411                         err = -ENOSYS;
 412                         goto error;
 413                 }
 414
 415                 err = skb_unclone(skb, GFP_ATOMIC);
 416                 if (unlikely(err))
 417                         goto error;
 418
 419                 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
 420         } else if (skb->ip_summed != CHECKSUM_PARTIAL)
 421                 skb->ip_summed = CHECKSUM_NONE;
 422
 423         skb->encapsulation = 1;
 424         return skb;
 425 error:
 426         kfree_skb(skb);
 427         return ERR_PTR(err);
 428 }
 429 #endif
 430
 431 static int lisp_send(struct vport *vport, struct sk_buff *skb)
 432 {
 433         struct ovs_key_ipv4_tunnel *tun_key;
 434         int network_offset = skb_network_offset(skb);
 435         struct rtable *rt;
 436         int min_headroom;
 437         __be32 saddr;
 438         __be16 df;
 439         int sent_len;
 440         int err;
 441
 442         if (unlikely(!OVS_CB(skb)->egress_tun_info)) {
 443                 err = -EINVAL;
 444                 goto error;
 445         }
 446
 447         tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
 448
 449         if (skb->protocol != htons(ETH_P_IP) &&
 450             skb->protocol != htons(ETH_P_IPV6)) {
 451                 err = 0;
 452                 goto error;
 453         }
 454
 455         /* Route lookup */
 456         saddr = tun_key->ipv4_src;
 457         rt = find_route(ovs_dp_get_net(vport->dp),
 458                         &saddr, tun_key->ipv4_dst,
 459                         IPPROTO_UDP, tun_key->ipv4_tos,
 460                         skb->mark);
 461         if (IS_ERR(rt)) {
 462                 err = PTR_ERR(rt);
 463                 goto error;
 464         }
 465
 466         min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
 467                         + sizeof(struct iphdr) + LISP_HLEN;
 468
 469         if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
 470                 int head_delta = SKB_DATA_ALIGN(min_headroom -
 471                                                 skb_headroom(skb) +
 472                                                 16);
 473
 474                 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
 475                                         0, GFP_ATOMIC);
 476                 if (unlikely(err))
 477                         goto err_free_rt;
 478         }
 479
 480         /* Reset l2 headers. */
 481         skb_pull(skb, network_offset);
 482         skb_reset_mac_header(skb);
 483         vlan_set_tci(skb, 0);
 484
 485         skb_reset_inner_headers(skb);
 486
 487         __skb_push(skb, LISP_HLEN);
 488         skb_reset_transport_header(skb);
 489
 490         lisp_build_header(vport, skb);
 491
 492         /* Offloading */
 493         skb = handle_offloads(skb);
 494         if (IS_ERR(skb)) {
 495                 err = PTR_ERR(skb);
 496                 skb = NULL;
 497                 goto err_free_rt;
 498         }
 499
 500         skb->ignore_df = 1;
 501
 502         ovs_skb_set_inner_protocol(skb, skb->protocol);
 503
 504         df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
 505         sent_len = iptunnel_xmit(skb->sk, rt, skb,
 506                              saddr, tun_key->ipv4_dst,
 507                              IPPROTO_UDP, tun_key->ipv4_tos,
 508                              tun_key->ipv4_ttl, df, false);
 509
 510         return sent_len > 0 ? sent_len + network_offset : sent_len;
 511
 512 err_free_rt:
 513         ip_rt_put(rt);
 514 error:
 515         kfree_skb(skb);
 516         return err;
 517 }
 518
 519 static const char *lisp_get_name(const struct vport *vport)
 520 {
 521         struct lisp_port *lisp_port = lisp_vport(vport);
 522         return lisp_port->name;
 523 }
 524
 525 static int lisp_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
 526                                     struct ovs_tunnel_info *egress_tun_info)
 527 {
 528         struct net *net = ovs_dp_get_net(vport->dp);
 529         struct lisp_port *lisp_port = lisp_vport(vport);
 530
 531         if (skb->protocol != htons(ETH_P_IP) &&
 532             skb->protocol != htons(ETH_P_IPV6)) {
 533                 return -EINVAL;
 534         }
 535
 536         /*
 537          * Get tp_src and tp_dst, refert to lisp_build_header().
 538          */
 539         return ovs_tunnel_get_egress_info(egress_tun_info, net,
 540                                           OVS_CB(skb)->egress_tun_info,
 541                                           IPPROTO_UDP, skb->mark,
 542                                           htons(get_src_port(net, skb)),
 543                                           lisp_port->dst_port);
 544 }
 545
 546 const struct vport_ops ovs_lisp_vport_ops = {
 547         .type                   = OVS_VPORT_TYPE_LISP,
 548         .create                 = lisp_tnl_create,
 549         .destroy                = lisp_tnl_destroy,
 550         .get_name               = lisp_get_name,
 551         .get_options            = lisp_get_options,
 552         .send                   = lisp_send,
 553         .get_egress_tun_info    = lisp_get_egress_tun_info,
 554 };