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                 if (skb->protocol == htons(ETH_P_IP)) {
 175                         struct iphdr *iph;
 176                         int size = (sizeof(iph->saddr) * 2) / sizeof(u32);
 177
 178                         iph = (struct iphdr *) skb_inner_network_header(skb);
 179                         hash = jhash2((const u32 *)&iph->saddr, size, 0);
 180                 } else if (skb->protocol == htons(ETH_P_IPV6)) {
 181                         struct ipv6hdr *ipv6hdr;
 182
 183                         ipv6hdr = (struct ipv6hdr *) skb_inner_network_header(skb);
 184                         hash = jhash2((const u32 *)&ipv6hdr->saddr,
 185                                       (sizeof(struct in6_addr) * 2) / sizeof(u32), 0);
 186                 } else {
 187                         pr_warn_once("LISP inner protocol is not IP when "
 188                                      "calculating hash.\n");
 189                 }
 190         }
 191
 192         inet_get_local_port_range(net, &low, &high);
 193         range = (high - low) + 1;
 194         return (((u64) hash * range) >> 32) + low;
 195 }
 196
 197 static void lisp_build_header(const struct vport *vport,
 198                               struct sk_buff *skb)
 199 {
 200         struct net *net = ovs_dp_get_net(vport->dp);
 201         struct lisp_port *lisp_port = lisp_vport(vport);
 202         struct udphdr *udph = udp_hdr(skb);
 203         struct lisphdr *lisph = (struct lisphdr *)(udph + 1);
 204         const struct ovs_key_ipv4_tunnel *tun_key;
 205
 206         tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
 207         udph->dest = lisp_port->dst_port;
 208         udph->source = htons(get_src_port(net, skb));
 209         udph->check = 0;
 210         udph->len = htons(skb->len - skb_transport_offset(skb));
 211
 212         lisph->nonce_present = 0;       /* We don't support echo nonce algorithm */
 213         lisph->locator_status_bits_present = 1; /* Set LSB */
 214         lisph->solicit_echo_nonce = 0;  /* No echo noncing */
 215         lisph->map_version_present = 0; /* No mapping versioning, nonce instead */
 216         lisph->instance_id_present = 1; /* Store the tun_id as Instance ID  */
 217         lisph->reserved_flags = 0;      /* Reserved flags, set to 0  */
 218
 219         lisph->u1.nonce[0] = 0;
 220         lisph->u1.nonce[1] = 0;
 221         lisph->u1.nonce[2] = 0;
 222
 223         tunnel_id_to_instance_id(tun_key->tun_id, &lisph->u2.word2.instance_id[0]);
 224         lisph->u2.word2.locator_status_bits = 1;
 225 }
 226
 227 /* Called with rcu_read_lock and BH disabled. */
 228 static int lisp_rcv(struct sock *sk, struct sk_buff *skb)
 229 {
 230         struct lisp_port *lisp_port;
 231         struct lisphdr *lisph;
 232         struct iphdr *iph, *inner_iph;
 233         struct ovs_tunnel_info tun_info;
 234         __be64 key;
 235         struct ethhdr *ethh;
 236         __be16 protocol;
 237
 238         lisp_port = lisp_find_port(dev_net(skb->dev), udp_hdr(skb)->dest);
 239         if (unlikely(!lisp_port))
 240                 goto error;
 241
 242         if (iptunnel_pull_header(skb, LISP_HLEN, 0))
 243                 goto error;
 244
 245         lisph = lisp_hdr(skb);
 246
 247         if (lisph->instance_id_present != 1)
 248                 key = 0;
 249         else
 250                 key = instance_id_to_tunnel_id(&lisph->u2.word2.instance_id[0]);
 251
 252         /* Save outer tunnel values */
 253         iph = ip_hdr(skb);
 254         ovs_flow_tun_info_init(&tun_info, iph,
 255                                udp_hdr(skb)->source, udp_hdr(skb)->dest,
 256                                key, TUNNEL_KEY, NULL, 0);
 257
 258         /* Drop non-IP inner packets */
 259         inner_iph = (struct iphdr *)(lisph + 1);
 260         switch (inner_iph->version) {
 261         case 4:
 262                 protocol = htons(ETH_P_IP);
 263                 break;
 264         case 6:
 265                 protocol = htons(ETH_P_IPV6);
 266                 break;
 267         default:
 268                 goto error;
 269         }
 270         skb->protocol = protocol;
 271
 272         /* Add Ethernet header */
 273         ethh = (struct ethhdr *)skb_push(skb, ETH_HLEN);
 274         memset(ethh, 0, ETH_HLEN);
 275         ethh->h_dest[0] = 0x02;
 276         ethh->h_source[0] = 0x02;
 277         ethh->h_proto = protocol;
 278
 279         ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN);
 280
 281         ovs_vport_receive(vport_from_priv(lisp_port), skb, &tun_info);
 282         goto out;
 283
 284 error:
 285         kfree_skb(skb);
 286 out:
 287         return 0;
 288 }
 289
 290 /* Arbitrary value.  Irrelevant as long as it's not 0 since we set the handler. */
 291 #define UDP_ENCAP_LISP 1
 292 static int lisp_socket_init(struct lisp_port *lisp_port, struct net *net)
 293 {
 294         struct sockaddr_in sin;
 295         int err;
 296
 297         err = sock_create_kern(AF_INET, SOCK_DGRAM, 0,
 298                                &lisp_port->lisp_rcv_socket);
 299         if (err)
 300                 goto error;
 301
 302         /* release net ref. */
 303         sk_change_net(lisp_port->lisp_rcv_socket->sk, net);
 304
 305         sin.sin_family = AF_INET;
 306         sin.sin_addr.s_addr = htonl(INADDR_ANY);
 307         sin.sin_port = lisp_port->dst_port;
 308
 309         err = kernel_bind(lisp_port->lisp_rcv_socket, (struct sockaddr *)&sin,
 310                           sizeof(struct sockaddr_in));
 311         if (err)
 312                 goto error_sock;
 313
 314         udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_type = UDP_ENCAP_LISP;
 315         udp_sk(lisp_port->lisp_rcv_socket->sk)->encap_rcv = lisp_rcv;
 316
 317         udp_encap_enable();
 318
 319         return 0;
 320
 321 error_sock:
 322         sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
 323 error:
 324         pr_warn("cannot register lisp protocol handler: %d\n", err);
 325         return err;
 326 }
 327
 328 static int lisp_get_options(const struct vport *vport, struct sk_buff *skb)
 329 {
 330         struct lisp_port *lisp_port = lisp_vport(vport);
 331
 332         if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(lisp_port->dst_port)))
 333                 return -EMSGSIZE;
 334         return 0;
 335 }
 336
 337 static void lisp_tnl_destroy(struct vport *vport)
 338 {
 339         struct lisp_port *lisp_port = lisp_vport(vport);
 340
 341         list_del_rcu(&lisp_port->list);
 342         /* Release socket */
 343         sk_release_kernel(lisp_port->lisp_rcv_socket->sk);
 344
 345         ovs_vport_deferred_free(vport);
 346 }
 347
 348 static struct vport *lisp_tnl_create(const struct vport_parms *parms)
 349 {
 350         struct net *net = ovs_dp_get_net(parms->dp);
 351         struct nlattr *options = parms->options;
 352         struct lisp_port *lisp_port;
 353         struct vport *vport;
 354         struct nlattr *a;
 355         int err;
 356         u16 dst_port;
 357
 358         if (!options) {
 359                 err = -EINVAL;
 360                 goto error;
 361         }
 362
 363         a = nla_find_nested(options, OVS_TUNNEL_ATTR_DST_PORT);
 364         if (a && nla_len(a) == sizeof(u16)) {
 365                 dst_port = nla_get_u16(a);
 366         } else {
 367                 /* Require destination port from userspace. */
 368                 err = -EINVAL;
 369                 goto error;
 370         }
 371
 372         /* Verify if we already have a socket created for this port */
 373         if (lisp_find_port(net, htons(dst_port))) {
 374                 err = -EEXIST;
 375                 goto error;
 376         }
 377
 378         vport = ovs_vport_alloc(sizeof(struct lisp_port),
 379                                 &ovs_lisp_vport_ops, parms);
 380         if (IS_ERR(vport))
 381                 return vport;
 382
 383         lisp_port = lisp_vport(vport);
 384         lisp_port->dst_port = htons(dst_port);
 385         strncpy(lisp_port->name, parms->name, IFNAMSIZ);
 386
 387         err = lisp_socket_init(lisp_port, net);
 388         if (err)
 389                 goto error_free;
 390
 391         list_add_tail_rcu(&lisp_port->list, &lisp_ports);
 392         return vport;
 393
 394 error_free:
 395         ovs_vport_free(vport);
 396 error:
 397         return ERR_PTR(err);
 398 }
 399
 400 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,12,0)
 401
 402 static void lisp_fix_segment(struct sk_buff *skb)
 403 {
 404         struct udphdr *udph = udp_hdr(skb);
 405
 406         udph->len = htons(skb->len - skb_transport_offset(skb));
 407 }
 408
 409 static int handle_offloads(struct sk_buff *skb)
 410 {
 411         if (skb_is_gso(skb))
 412                 OVS_GSO_CB(skb)->fix_segment = lisp_fix_segment;
 413         else if (skb->ip_summed != CHECKSUM_PARTIAL)
 414                 skb->ip_summed = CHECKSUM_NONE;
 415         return 0;
 416 }
 417 #else
 418 static int handle_offloads(struct sk_buff *skb)
 419 {
 420         if (skb->encapsulation && skb_is_gso(skb)) {
 421                 kfree_skb(skb);
 422                 return -ENOSYS;
 423         }
 424
 425         if (skb_is_gso(skb)) {
 426                 int err = skb_unclone(skb, GFP_ATOMIC);
 427                 if (unlikely(err))
 428                         return err;
 429
 430                 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
 431         } else if (skb->ip_summed != CHECKSUM_PARTIAL)
 432                 skb->ip_summed = CHECKSUM_NONE;
 433
 434         skb->encapsulation = 1;
 435         return 0;
 436 }
 437 #endif
 438
 439 static int lisp_send(struct vport *vport, struct sk_buff *skb)
 440 {
 441         struct ovs_key_ipv4_tunnel *tun_key;
 442         int network_offset = skb_network_offset(skb);
 443         struct rtable *rt;
 444         int min_headroom;
 445         __be32 saddr;
 446         __be16 df;
 447         int sent_len;
 448         int err;
 449
 450         if (unlikely(!OVS_CB(skb)->egress_tun_info))
 451                 return -EINVAL;
 452
 453         tun_key = &OVS_CB(skb)->egress_tun_info->tunnel;
 454
 455         if (skb->protocol != htons(ETH_P_IP) &&
 456             skb->protocol != htons(ETH_P_IPV6)) {
 457                 kfree_skb(skb);
 458                 return 0;
 459         }
 460
 461         /* Route lookup */
 462         saddr = tun_key->ipv4_src;
 463         rt = find_route(ovs_dp_get_net(vport->dp),
 464                         &saddr, tun_key->ipv4_dst,
 465                         IPPROTO_UDP, tun_key->ipv4_tos,
 466                         skb->mark);
 467         if (IS_ERR(rt)) {
 468                 err = PTR_ERR(rt);
 469                 goto error;
 470         }
 471
 472         min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
 473                         + sizeof(struct iphdr) + LISP_HLEN;
 474
 475         if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) {
 476                 int head_delta = SKB_DATA_ALIGN(min_headroom -
 477                                                 skb_headroom(skb) +
 478                                                 16);
 479
 480                 err = pskb_expand_head(skb, max_t(int, head_delta, 0),
 481                                         0, GFP_ATOMIC);
 482                 if (unlikely(err))
 483                         goto err_free_rt;
 484         }
 485
 486         /* Reset l2 headers. */
 487         skb_pull(skb, network_offset);
 488         skb_reset_mac_header(skb);
 489         vlan_set_tci(skb, 0);
 490
 491         skb_reset_inner_headers(skb);
 492
 493         __skb_push(skb, LISP_HLEN);
 494         skb_reset_transport_header(skb);
 495
 496         lisp_build_header(vport, skb);
 497
 498         /* Offloading */
 499         err = handle_offloads(skb);
 500         if (err)
 501                 goto err_free_rt;
 502
 503         skb->ignore_df = 1;
 504
 505         df = tun_key->tun_flags & TUNNEL_DONT_FRAGMENT ? htons(IP_DF) : 0;
 506         sent_len = iptunnel_xmit(skb->sk, rt, skb,
 507                              saddr, tun_key->ipv4_dst,
 508                              IPPROTO_UDP, tun_key->ipv4_tos,
 509                              tun_key->ipv4_ttl, df, false);
 510
 511         return sent_len > 0 ? sent_len + network_offset : sent_len;
 512
 513 err_free_rt:
 514         ip_rt_put(rt);
 515 error:
 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 };