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/xfrm.h>
  35
  36 #include "datapath.h"
  37 #include "gso.h"
  38 #include "vport.h"
  39
  40 /*
  41  *  LISP encapsulation header:
  42  *
  43  *  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  44  *  |N|L|E|V|I|flags|            Nonce/Map-Version                  |
  45  *  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  46  *  |                 Instance ID/Locator Status Bits               |
  47  *  +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
  48  *
  49  */
  50
  51 /**
  52  * struct lisphdr - LISP header
  53  * @nonce_present: Flag indicating the presence of a 24 bit nonce value.
  54  * @locator_status_bits_present: Flag indicating the presence of Locator Status
  55  *                               Bits (LSB).
  56  * @solicit_echo_nonce: Flag indicating the use of the echo noncing mechanism.
  57  * @map_version_present: Flag indicating the use of mapping versioning.
  58  * @instance_id_present: Flag indicating the presence of a 24 bit Instance ID.
  59  * @reserved_flags: 3 bits reserved for future flags.
  60  * @nonce: 24 bit nonce value.
  61  * @map_version: 24 bit mapping version.
  62  * @locator_status_bits: Locator Status Bits: 32 bits when instance_id_present
  63  *                       is not set, 8 bits when it is.
  64  * @instance_id: 24 bit Instance ID
  65  */
  66 struct lisphdr {
  67 #ifdef __LITTLE_ENDIAN_BITFIELD
  68         __u8 reserved_flags:3;
  69         __u8 instance_id_present:1;
  70         __u8 map_version_present:1;
  71         __u8 solicit_echo_nonce:1;
  72         __u8 locator_status_bits_present:1;
  73         __u8 nonce_present:1;
  74 #else
  75         __u8 nonce_present:1;
  76         __u8 locator_status_bits_present:1;
  77         __u8 solicit_echo_nonce:1;
  78         __u8 map_version_present:1;
  79         __u8 instance_id_present:1;
  80         __u8 reserved_flags:3;
  81 #endif
  82         union {
  83                 __u8 nonce[3];
  84                 __u8 map_version[3];
  85         } u1;
  86         union {
  87                 __be32 locator_status_bits;
  88                 struct {
  89                         __u8 instance_id[3];
  90                         __u8 locator_status_bits;
  91                 } word2;
  92         } u2;
  93 };
  94
  95 #define LISP_HLEN (sizeof(struct udphdr) + sizeof(struct lisphdr))
  96
  97 /**
  98  * struct lisp_port - Keeps track of open UDP ports
  99  * @dst_port: lisp UDP port no.
 100  * @list: list element in @lisp_ports.
 101  * @lisp_rcv_socket: The socket created for this port number.
 102  * @name: vport name.
 103  */
 104 struct lisp_port {
 105         __be16 dst_port;
 106         struct list_head list;
 107         struct socket *lisp_rcv_socket;
 108         char name[IFNAMSIZ];
 109 };
 110
 111 static LIST_HEAD(lisp_ports);
 112
 113 static inline struct lisp_port *lisp_vport(const struct vport *vport)
 114 {
 115         return vport_priv(vport);
 116 }
 117
 118 static struct lisp_port *lisp_find_port(struct net *net, __be16 port)
 119 {
 120         struct lisp_port *lisp_port;
 121
 122         list_for_each_entry_rcu(lisp_port, &lisp_ports, list) {
 123                 if (lisp_port->dst_port == port &&
 124                         net_eq(sock_net(lisp_port->lisp_rcv_socket->sk), net))
 125                         return lisp_port;
 126         }
 127
 128         return NULL;
 129 }
 130
 131 static inline struct lisphdr *lisp_hdr(const struct sk_buff *skb)
 132 {
 133         return (struct lisphdr *)(udp_hdr(skb) + 1);
 134 }
 135
 136 /* Convert 64 bit tunnel ID to 24 bit Instance ID. */
 137 static void tunnel_id_to_instance_id(__be64 tun_id, __u8 *iid)
 138 {
 139
 140 #ifdef __BIG_ENDIAN
 141         iid[0] = (__force __u8)(tun_id >> 16);
 142         iid[1] = (__force __u8)(tun_id >> 8);
 143         iid[2] = (__force __u8)tun_id;
 144 #else
 145         iid[0] = (__force __u8)((__force u64)tun_id >> 40);
 146         iid[1] = (__force __u8)((__force u64)tun_id >> 48);
 147         iid[2] = (__force __u8)((__force u64)tun_id >> 56);
 148 #endif
 149 }
 150
 151 /* Convert 24 bit Instance ID to 64 bit tunnel ID. */
 152 static __be64 instance_id_to_tunnel_id(__u8 *iid)
 153 {
 154 #ifdef __BIG_ENDIAN
 155         return (iid[0] << 16) | (iid[1] << 8) | iid[2];
 156 #else
 157         return (__force __be64)(((__force u64)iid[0] << 40) |
 158                                 ((__force u64)iid[1] << 48) |
 159                                 ((__force u64)iid[2] << 56));
 160 #endif
 161 }
 162
 163 /* Compute source UDP port for outgoing packet.
 164  * Currently we use the flow hash.
 165  */
 166 static u16 get_src_port(struct net *net, struct sk_buff *skb)
 167 {
 168         u32 hash = skb_get_hash(skb);
 169         unsigned int range;
 170         int high;
 171         int low;
 172
 173         if (!hash) {
 174                 struct sw_flow_key *pkt_key = OVS_CB(skb)->pkt_key;
 175
 176                 if (skb->protocol == htons(ETH_P_IP))
 177                         hash = jhash2((const u32 *)&pkt_key->ipv4.addr,
 178                                    sizeof(pkt_key->ipv4.addr) / sizeof(u32), 0);
 179                 else if (skb->protocol == htons(ETH_P_IPV6))
 180                         hash = jhash2((const u32 *)&pkt_key->ipv6.addr,
 181                                    sizeof(pkt_key->ipv6.addr) / sizeof(u32), 0);
 182                 else
 183                         pr_warn_once("LISP inner protocol is not IP when "
 184                                      "calculating hash.\n");
 185         }
 186
 187         inet_get_local_port_range(net, &low, &high);
 188         range = (high - low) + 1;
 189         return (((u64) hash * range) >> 32) + low;
 190 }
 191
 192 static void lisp_build_header(const struct vport *vport,
 193                               struct sk_buff *skb)
 194 {
 195         struct net *net = ovs_dp_get_net(vport->dp);
 196         struct lisp_port *lisp_port = lisp_vport(vport);
 197         struct udphdr *udph = udp_hdr(skb);
 198         struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
 199         const struct ovs_key_ipv4_tunnel *tun_key = &OVS_CB(skb)->tun_info->tunnel;
 200
 201         udph->dest = lisp_port->dst_port;
 202         udph->source = htons(get_src_port(net, skb));
 203         udph->check = 0;
 204         udph->len = htons(skb->len - skb_transport_offset(skb));
 205
 206         lisph->nonce_present = 0;       /* We don't support echo nonce algorithm */
 207         lisph->locator_status_bits_present = 1; /* Set LSB */
 208         lisph->solicit_echo_nonce = 0;  /* No echo noncing */
 209         lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
 210         lisph->instance_id_present = 1; /* Store the tun_id as Instance ID  */
 211         lisph->reserved_flags = 0;      /* Reserved flags, set to 0  */
 212
 213         lisph->u1.nonce[0] = 0;
 214         lisph->u1.nonce[1] = 0;
 215         lisph->u1.nonce[2] = 0;
 216
 217         tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
 218         lisph->u2.word2.locator_status_bits = 1;
 219 }
 220
 221 /* Called with rcu_read_lock and BH disabled. */
 222 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 223 {
 224         struct lisp_port *lisp_port;
 225         struct lisphdr *lisph;
 226         struct iphdr *iph, *inner_iph;
 227         struct ovs_tunnel_info tun_info;
 228         __be64 key;
 229         struct ethhdr *ethh;
 230         __be16 protocol;
 231
 232         lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
 233         if (unlikely(!lisp_port))
 234                 goto error;
 235
 236         if (iptunnel_pull_header(skb, LISP_HLEN, 0))
 237                 goto error;
 238
 239         lisph = lisp_hdr(skb);
 240
 241         if (lisph->instance_id_present != 1)
 242                 key = 0;
 243         else
 244                 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
 245
 246         /* Save outer tunnel values */
 247         iph = ip_hdr(skb);
 248         ovs_flow_tun_info_init(&tun_info, iph, key, TUNNEL_KEY, NULL, 0);
 249
 250         /* Drop non-IP inner packets */
 251         inner_iph = (struct iphdr *)(lisph + 1);
 252         switch (inner_iph->version) {
 253         case 4:
 254                 protocol = htons(ETH_P_IP);
 255                 break;
 256         case 6:
 257                 protocol = htons(ETH_P_IPV6);
 258                 break;
 259         default:
 260                 goto error;
 261         }
 262         skb->protocol = protocol;
 263
 264         /* Add Ethernet header */
 265         ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
 266         memset(ethh, 0, ETH_HLEN);
 267         ethh->h_dest[0] = 0x02;
 268         ethh->h_source[0] = 0x02;
 269         ethh->h_proto = protocol;
 270
 271         ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
 272
 273         ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_info);
 274         goto out;
 275
 276 error:
 277         kfree_skb(skb);
 278 out:
 279         return 0;
 280 }
 281
 282 /* Arbitrary value.  Irrelevant as long as it's not 0 since we set the handler. */
 283 #define UDP_ENCAP_LISP 1
 284 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
 285 {
 286         struct sockaddr_in sin;
 287         int err;
 288
 289         err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
 290                                &lisp_port->lisp_rcv_socket);
 291         if (err)
 292                 goto error;
 293
 294         /* release net ref. */
 295         sk_change_net(lisp_port->lisp_rcv_socket->sk, net);
 296
 297         sin.sin_family = AF_INET;
 298         sin.sin_addr.s_addr = htonl(INADDR_ANY);
 299         sin.sin_port = lisp_port->dst_port;
 300
 301         err = kernel_bind(lisp_port->lisp_rcv_socket, (struct sockaddr *)&sin,
 302                           sizeof(struct sockaddr_in));
 303         if (err)
 304                 goto error_sock;
 305
 306         udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_type = UDP_ENCAP_LISP;
 307         udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_rcv = lisp_rcv;
 308
 309         udp_encap_enable();
 310
 311         return 0;
 312
 313 error_sock:
 314         sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
 315 error:
 316         pr_warn("cannot register lisp protocol handler: %d\n", err);
 317         return err;
 318 }
 319
 320 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
 321 {
 322         struct lisp_port *lisp_port = lisp_vport(vport);
 323
 324         if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
 325                 return -EMSGSIZE;
 326         return 0;
 327 }
 328
 329 static void lisp_tnl_destroy(struct vport *vport)
 330 {
 331         struct lisp_port *lisp_port = lisp_vport(vport);
 332
 333         list_del_rcu(&lisp_port->list);
 334         /* Release socket */
 335         sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
 336
 337         ovs_vport_deferred_free(vport);
 338 }
 339
 340 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
 341 {
 342         struct net *net = ovs_dp_get_net(parms->dp);
 343         struct nlattr *options = parms->options;
 344         struct lisp_port *lisp_port;
 345         struct vport *vport;
 346         struct nlattr *a;
 347         int err;
 348         u16 dst_port;
 349
 350         if (!options) {
 351                 err = -EINVAL;
 352                 goto error;
 353         }
 354
 355         a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
 356         if (a && nla_len(a) == sizeof(u16)) {
 357                 dst_port = nla_get_u16(a);
 358         } else {
 359                 /* Require destination port from userspace. */
 360                 err = -EINVAL;
 361                 goto error;
 362         }
 363
 364         /* Verify if we already have a socket created for this port */
 365         if (lisp_find_port(net, htons(dst_port))) {
 366                 err = -EEXIST;
 367                 goto error;
 368         }
 369
 370         vport = ovs_vport_alloc(sizeof(struct lisp_port),
 371                                 &ovs_lisp_vport_ops, parms);
 372         if (IS_ERR(vport))
 373                 return vport;
 374
 375         lisp_port = lisp_vport(vport);
 376         lisp_port->dst_port = htons(dst_port);
 377         strncpy(lisp_port->name, parms->name, IFNAMSIZ);
 378
 379         err = lisp_socket_init(lisp_port, net);
 380         if (err)
 381                 goto error_free;
 382
 383         list_add_tail_rcu(&lisp_port->list, &lisp_ports);
 384         return vport;
 385
 386 error_free:
 387         ovs_vport_free(vport);
 388 error:
 389         return ERR_PTR(err);
 390 }
 391
 392 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,12,0)
 393
 394 static void lisp_fix_segment(struct sk_buff *skb)
 395 {
 396         struct udphdr *udph = udp_hdr(skb);
 397
 398         udph->len = htons(skb->len - skb_transport_offset(skb));
 399 }
 400
 401 static int handle_offloads(struct sk_buff *skb)
 402 {
 403         if (skb_is_gso(skb))
 404                 OVS_GSO_CB(skb)->fix_segment = lisp_fix_segment;
 405         else if (skb->ip_summed != CHECKSUM_PARTIAL)
 406                 skb->ip_summed = CHECKSUM_NONE;
 407         return 0;
 408 }
 409 #else
 410 static int handle_offloads(struct sk_buff *skb)
 411 {
 412         if (skb_is_gso(skb)) {
 413                 int err = skb_unclone(skb, GFP_ATOMIC);
 414                 if (unlikely(err))
 415                         return err;
 416
 417                 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
 418         } else if (skb->ip_summed != CHECKSUM_PARTIAL)
 419                 skb->ip_summed = CHECKSUM_NONE;
 420
 421         skb->encapsulation = 1;
 422         return 0;
 423 }
 424 #endif
 425
 426 static int lisp_send(struct vport *vport, struct sk_buff *skb)
 427 {
 428         struct ovs_key_ipv4_tunnel *tun_key;
 429         int network_offset = skb_network_offset(skb);
 430         struct rtable *rt;
 431         int min_headroom;
 432         __be32 saddr;
 433         __be16 df;
 434         int sent_len;
 435         int err;
 436
 437         if (unlikely(!OVS_CB(skb)->tun_info))
 438                 return -EINVAL;
 439
 440         tun_key = &OVS_CB(skb)->tun_info->tunnel;
 441
 442         if (skb->protocol != htons(ETH_P_IP) &&
 443             skb->protocol != htons(ETH_P_IPV6)) {
 444                 kfree_skb(skb);
 445                 return 0;
 446         }
 447
 448         /* Route lookup */
 449         saddr = tun_key->ipv4_src;
 450         rt = find_route(ovs_dp_get_net(vport->dp),
 451                         &saddr, tun_key->ipv4_dst,
 452                         IPPROTO_UDP, tun_key->ipv4_tos,
 453                         skb->mark);
 454         if (IS_ERR(rt)) {
 455                 err = PTR_ERR(rt);
 456                 goto error;
 457         }
 458
 459         min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
 460                         + sizeof(struct iphdr) + LISP_HLEN;
 461
 462         if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
 463                 int head_delta = SKB_DATA_ALIGN(min_headroom -
 464                                                 skb_headroom(skb) +
 465                                                 16);
 466
 467                 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
 468                                         0, GFP_ATOMIC);
 469                 if (unlikely(err))
 470                         goto err_free_rt;
 471         }
 472
 473         /* Reset l2 headers. */
 474         skb_pull(skb, network_offset);
 475         skb_reset_mac_header(skb);
 476         vlan_set_tci(skb, 0);
 477
 478         skb_reset_inner_headers(skb);
 479
 480         __skb_push(skb, LISP_HLEN);
 481         skb_reset_transport_header(skb);
 482
 483         lisp_build_header(vport, skb);
 484
 485         /* Offloading */
 486         err = handle_offloads(skb);
 487         if (err)
 488                 goto err_free_rt;
 489
 490         skb->local_df = 1;
 491
 492         df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
 493         sent_len = iptunnel_xmit(rt, skb,
 494                              saddr, tun_key->ipv4_dst,
 495                              IPPROTO_UDP, tun_key->ipv4_tos,
 496                              tun_key->ipv4_ttl, df, false);
 497
 498         return sent_len > 0 ? sent_len + network_offset : sent_len;
 499
 500 err_free_rt:
 501         ip_rt_put(rt);
 502 error:
 503         return err;
 504 }
 505
 506 static const char *lisp_get_name(const struct vport *vport)
 507 {
 508         struct lisp_port *lisp_port = lisp_vport(vport);
 509         return lisp_port->name;
 510 }
 511
 512 const struct vport_ops ovs_lisp_vport_ops = {
 513         .type           = OVS_VPORT_TYPE_LISP,
 514         .create         = lisp_tnl_create,
 515         .destroy        = lisp_tnl_destroy,
 516         .get_name       = lisp_get_name,
 517         .get_options    = lisp_get_options,
 518         .send           = lisp_send,
 519 };