ofproto/tunnel.c

   1 /* Copyright (c) 2013, 2014, 2015, 2017 Nicira, Inc.
   2  *
   3  * Licensed under the Apache License, Version 2.0 (the "License");
   4  * you may not use this file except in compliance with the License.
   5  * You may obtain a copy of the License at:
   6  *
   7  *     http://www.apache.org/licenses/LICENSE-2.0
   8  *
   9  * Unless required by applicable law or agreed to in writing, software
  10  * distributed under the License is distributed on an "AS IS" BASIS,
  11  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12  * See the License for the specific language governing permissions and
  13  * limitations under the License. */
  14
  15 #include <config.h>
  16 #include "tunnel.h"
  17
  18 #include <errno.h>
  19
  20 #include "byte-order.h"
  21 #include "connectivity.h"
  22 #include "csum.h"
  23 #include "dpif.h"
  24 #include "openvswitch/dynamic-string.h"
  25 #include "fat-rwlock.h"
  26 #include "hash.h"
  27 #include "openvswitch/hmap.h"
  28 #include "netdev.h"
  29 #include "odp-util.h"
  30 #include "openvswitch/ofpbuf.h"
  31 #include "packets.h"
  32 #include "route-table.h"
  33 #include "seq.h"
  34 #include "smap.h"
  35 #include "socket-util.h"
  36 #include "tnl-ports.h"
  37 #include "tunnel.h"
  38 #include "openvswitch/vlog.h"
  39 #include "unaligned.h"
  40 #include "unixctl.h"
  41 #include "ofproto-dpif.h"
  42 #include "netdev-vport.h"
  43
  44 VLOG_DEFINE_THIS_MODULE(tunnel);
  45
  46 struct tnl_match {
  47     ovs_be64 in_key;
  48     struct in6_addr ipv6_src;
  49     struct in6_addr ipv6_dst;
  50     odp_port_t odp_port;
  51     bool in_key_flow;
  52     bool ip_src_flow;
  53     bool ip_dst_flow;
  54     enum netdev_pt_mode pt_mode;
  55 };
  56
  57 struct tnl_port {
  58     struct hmap_node ofport_node;
  59     struct hmap_node match_node;
  60
  61     const struct ofport_dpif *ofport;
  62     uint64_t change_seq;
  63     struct netdev *netdev;
  64
  65     struct tnl_match match;
  66 };
  67
  68 static struct fat_rwlock rwlock;
  69
  70 /* Tunnel matches.
  71  *
  72  * This module maps packets received over tunnel protocols to vports.  The
  73  * tunnel protocol and, for some protocols, tunnel-specific information (e.g.,
  74  * for VXLAN, the UDP destination port number) are always use as part of the
  75  * mapping.  Which other fields are used for the mapping depends on the vports
  76  * themselves (the parenthesized notations refer to "struct tnl_match" fields):
  77  *
  78  *     - in_key: A vport may match a specific tunnel ID (in_key_flow == false)
  79  *       or arrange for the tunnel ID to be matched as tunnel.tun_id in the
  80  *       OpenFlow flow (in_key_flow == true).
  81  *
  82  *     - ip_dst: A vport may match a specific destination IP address
  83  *       (ip_dst_flow == false) or arrange for the destination IP to be matched
  84  *       as tunnel.ip_dst in the OpenFlow flow (ip_dst_flow == true).
  85  *
  86  *     - ip_src: A vport may match a specific IP source address (ip_src_flow ==
  87  *       false, ip_src != 0), wildcard all source addresses (ip_src_flow ==
  88  *       false, ip_src == 0), or arrange for the IP source address to be
  89  *       handled in the OpenFlow flow table (ip_src_flow == true).
  90  *
  91  * Thus, there are 2 * 2 * 3 == 12 possible ways a vport can match against a
  92  * tunnel packet.  We number the possibilities for each field in increasing
  93  * order as listed in each bullet above.  We order the 12 overall combinations
  94  * in lexicographic order considering in_key first, then ip_dst, then
  95  * ip_src. */
  96 #define N_MATCH_TYPES (2 * 2 * 3)
  97
  98 /* The three possibilities (see above) for vport ip_src matches. */
  99 enum ip_src_type {
 100     IP_SRC_CFG,             /* ip_src must equal configured address. */
 101     IP_SRC_ANY,             /* Any ip_src is acceptable. */
 102     IP_SRC_FLOW             /* ip_src is handled in flow table. */
 103 };
 104
 105 /* Each hmap contains "struct tnl_port"s.
 106  * The index is a combination of how each of the fields listed under "Tunnel
 107  * matches" above matches, see the final paragraph for ordering. */
 108 static struct hmap *tnl_match_maps[N_MATCH_TYPES] OVS_GUARDED_BY(rwlock);
 109 static struct hmap **tnl_match_map(const struct tnl_match *);
 110
 111 static struct hmap ofport_map__ = HMAP_INITIALIZER(&ofport_map__);
 112 static struct hmap *ofport_map OVS_GUARDED_BY(rwlock) = &ofport_map__;
 113
 114 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
 115 static struct vlog_rate_limit dbg_rl = VLOG_RATE_LIMIT_INIT(60, 60);
 116
 117 static struct tnl_port *tnl_find(const struct flow *) OVS_REQ_RDLOCK(rwlock);
 118 static struct tnl_port *tnl_find_exact(struct tnl_match *, struct hmap *)
 119     OVS_REQ_RDLOCK(rwlock);
 120 static struct tnl_port *tnl_find_ofport(const struct ofport_dpif *)
 121     OVS_REQ_RDLOCK(rwlock);
 122
 123 static uint32_t tnl_hash(struct tnl_match *);
 124 static void tnl_match_fmt(const struct tnl_match *, struct ds *);
 125 static char *tnl_port_to_string(const struct tnl_port *)
 126     OVS_REQ_RDLOCK(rwlock);
 127 static void tnl_port_format(const struct tnl_port *, struct ds *)
 128     OVS_REQ_RDLOCK(rwlock);
 129 static void tnl_port_mod_log(const struct tnl_port *, const char *action)
 130     OVS_REQ_RDLOCK(rwlock);
 131 static const char *tnl_port_get_name(const struct tnl_port *)
 132     OVS_REQ_RDLOCK(rwlock);
 133 static void tnl_port_del__(const struct ofport_dpif *, odp_port_t)
 134     OVS_REQ_WRLOCK(rwlock);
 135
 136 static unixctl_cb_func tnl_unixctl_list;
 137
 138 void
 139 ofproto_tunnel_init(void)
 140 {
 141     static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
 142
 143     if (ovsthread_once_start(&once)) {
 144         fat_rwlock_init(&rwlock);
 145         unixctl_command_register("ofproto/list-tunnels", "", 0, 0,
 146                                  tnl_unixctl_list, NULL);
 147         ovsthread_once_done(&once);
 148     }
 149 }
 150
 151 static bool
 152 tnl_port_add__(const struct ofport_dpif *ofport, const struct netdev *netdev,
 153                odp_port_t odp_port, bool warn, bool native_tnl, const char name[])
 154     OVS_REQ_WRLOCK(rwlock)
 155 {
 156     const struct netdev_tunnel_config *cfg;
 157     struct tnl_port *existing_port;
 158     struct tnl_port *tnl_port;
 159     struct hmap **map;
 160
 161     cfg = netdev_get_tunnel_config(netdev);
 162     ovs_assert(cfg);
 163
 164     tnl_port = xzalloc(sizeof *tnl_port);
 165     tnl_port->ofport = ofport;
 166     tnl_port->netdev = netdev_ref(netdev);
 167     tnl_port->change_seq = netdev_get_change_seq(tnl_port->netdev);
 168
 169     tnl_port->match.in_key = cfg->in_key;
 170     tnl_port->match.ipv6_src = cfg->ipv6_src;
 171     tnl_port->match.ipv6_dst = cfg->ipv6_dst;
 172     tnl_port->match.ip_src_flow = cfg->ip_src_flow;
 173     tnl_port->match.ip_dst_flow = cfg->ip_dst_flow;
 174     tnl_port->match.in_key_flow = cfg->in_key_flow;
 175     tnl_port->match.odp_port = odp_port;
 176     tnl_port->match.pt_mode = netdev_get_pt_mode(netdev);
 177
 178     map = tnl_match_map(&tnl_port->match);
 179     existing_port = tnl_find_exact(&tnl_port->match, *map);
 180     if (existing_port) {
 181         if (warn) {
 182             struct ds ds = DS_EMPTY_INITIALIZER;
 183             tnl_match_fmt(&tnl_port->match, &ds);
 184             VLOG_WARN("%s: attempting to add tunnel port with same config as "
 185                       "port '%s' (%s)", tnl_port_get_name(tnl_port),
 186                       tnl_port_get_name(existing_port), ds_cstr(&ds));
 187             ds_destroy(&ds);
 188         }
 189         netdev_close(tnl_port->netdev);
 190         free(tnl_port);
 191         return false;
 192     }
 193
 194     hmap_insert(ofport_map, &tnl_port->ofport_node, hash_pointer(ofport, 0));
 195
 196     if (!*map) {
 197         *map = xmalloc(sizeof **map);
 198         hmap_init(*map);
 199     }
 200     hmap_insert(*map, &tnl_port->match_node, tnl_hash(&tnl_port->match));
 201     tnl_port_mod_log(tnl_port, "adding");
 202
 203     if (native_tnl) {
 204         const char *type;
 205
 206         type = netdev_get_type(netdev);
 207         tnl_port_map_insert(odp_port, cfg->dst_port, name, type);
 208
 209     }
 210     return true;
 211 }
 212
 213 /* Adds 'ofport' to the module with datapath port number 'odp_port'. 'ofport's
 214  * must be added before they can be used by the module. 'ofport' must be a
 215  * tunnel.
 216  *
 217  * Returns 0 if successful, otherwise a positive errno value. */
 218 int
 219 tnl_port_add(const struct ofport_dpif *ofport, const struct netdev *netdev,
 220              odp_port_t odp_port, bool native_tnl, const char name[])
 221     OVS_EXCLUDED(rwlock)
 222 {
 223     bool ok;
 224
 225     fat_rwlock_wrlock(&rwlock);
 226     ok = tnl_port_add__(ofport, netdev, odp_port, true, native_tnl, name);
 227     fat_rwlock_unlock(&rwlock);
 228
 229     return ok ? 0 : EEXIST;
 230 }
 231
 232 /* Checks if the tunnel represented by 'ofport' reconfiguration due to changes
 233  * in its netdev_tunnel_config. If it does, returns true. Otherwise, returns
 234  * false. 'new_odp_port' should be the port number coming from 'ofport' that
 235  * is passed to tnl_port_add__(). 'old_odp_port' should be the port number
 236  * that is passed to tnl_port_del__(). */
 237 bool
 238 tnl_port_reconfigure(const struct ofport_dpif *ofport,
 239                      const struct netdev *netdev, odp_port_t new_odp_port,
 240                      odp_port_t old_odp_port, bool native_tnl,
 241                      const char name[])
 242     OVS_EXCLUDED(rwlock)
 243 {
 244     struct tnl_port *tnl_port;
 245     bool changed = false;
 246
 247     fat_rwlock_wrlock(&rwlock);
 248     tnl_port = tnl_find_ofport(ofport);
 249     if (!tnl_port) {
 250         changed = tnl_port_add__(ofport, netdev, new_odp_port, false,
 251                                  native_tnl, name);
 252     } else if (tnl_port->netdev != netdev
 253                || tnl_port->match.odp_port != new_odp_port
 254                || tnl_port->change_seq != netdev_get_change_seq(tnl_port->netdev)) {
 255         VLOG_DBG("reconfiguring %s", tnl_port_get_name(tnl_port));
 256         tnl_port_del__(ofport, old_odp_port);
 257         tnl_port_add__(ofport, netdev, new_odp_port, true, native_tnl, name);
 258         changed = true;
 259     }
 260     fat_rwlock_unlock(&rwlock);
 261     return changed;
 262 }
 263
 264 static void
 265 tnl_port_del__(const struct ofport_dpif *ofport, odp_port_t odp_port)
 266     OVS_REQ_WRLOCK(rwlock)
 267 {
 268     struct tnl_port *tnl_port;
 269
 270     if (!ofport) {
 271         return;
 272     }
 273
 274     tnl_port = tnl_find_ofport(ofport);
 275     if (tnl_port) {
 276         struct hmap **map;
 277
 278         tnl_port_map_delete(odp_port, netdev_get_type(tnl_port->netdev));
 279         tnl_port_mod_log(tnl_port, "removing");
 280         map = tnl_match_map(&tnl_port->match);
 281         hmap_remove(*map, &tnl_port->match_node);
 282         if (hmap_is_empty(*map)) {
 283             hmap_destroy(*map);
 284             free(*map);
 285             *map = NULL;
 286         }
 287         hmap_remove(ofport_map, &tnl_port->ofport_node);
 288         netdev_close(tnl_port->netdev);
 289         free(tnl_port);
 290     }
 291 }
 292
 293 /* Removes 'ofport' from the module. */
 294 void
 295 tnl_port_del(const struct ofport_dpif *ofport, odp_port_t odp_port)
 296     OVS_EXCLUDED(rwlock)
 297 {
 298     fat_rwlock_wrlock(&rwlock);
 299     tnl_port_del__(ofport, odp_port);
 300     fat_rwlock_unlock(&rwlock);
 301 }
 302
 303 /* Looks in the table of tunnels for a tunnel matching the metadata in 'flow'.
 304  * Returns the 'ofport' corresponding to the new in_port, or a null pointer if
 305  * none is found.
 306  *
 307  * Callers should verify that 'flow' needs to be received by calling
 308  * tnl_port_should_receive() before this function. */
 309 const struct ofport_dpif *
 310 tnl_port_receive(const struct flow *flow) OVS_EXCLUDED(rwlock)
 311 {
 312     const struct ofport_dpif *ofport;
 313     struct tnl_port *tnl_port;
 314
 315     fat_rwlock_rdlock(&rwlock);
 316     tnl_port = tnl_find(flow);
 317     ofport = tnl_port ? tnl_port->ofport : NULL;
 318     if (!tnl_port) {
 319         if (!VLOG_DROP_WARN(&rl)) {
 320             char *flow_str = flow_to_string(flow, NULL);
 321             VLOG_WARN("receive tunnel port not found (%s)", flow_str);
 322             free(flow_str);
 323         }
 324         goto out;
 325     }
 326
 327     if (!VLOG_DROP_DBG(&dbg_rl)) {
 328         char *flow_str = flow_to_string(flow, NULL);
 329         char *tnl_str = tnl_port_to_string(tnl_port);
 330         VLOG_DBG("tunnel port %s receive from flow %s", tnl_str, flow_str);
 331         free(tnl_str);
 332         free(flow_str);
 333     }
 334
 335 out:
 336     fat_rwlock_unlock(&rwlock);
 337     return ofport;
 338 }
 339
 340 /* Should be called at the beginning of action translation to initialize
 341  * wildcards and perform any actions based on receiving on tunnel port.
 342  *
 343  * Returns false if the packet must be dropped. */
 344 bool
 345 tnl_process_ecn(struct flow *flow)
 346 {
 347     if (!tnl_port_should_receive(flow)) {
 348         return true;
 349     }
 350
 351     if (is_ip_any(flow) && IP_ECN_is_ce(flow->tunnel.ip_tos)) {
 352         if ((flow->nw_tos & IP_ECN_MASK) == IP_ECN_NOT_ECT) {
 353             VLOG_WARN_RL(&rl, "dropping tunnel packet marked ECN CE"
 354                          " but is not ECN capable");
 355             return false;
 356         }
 357
 358         /* Set the ECN CE value in the tunneled packet. */
 359         flow->nw_tos |= IP_ECN_CE;
 360     }
 361
 362     return true;
 363 }
 364
 365 void
 366 tnl_wc_init(struct flow *flow, struct flow_wildcards *wc)
 367 {
 368     if (tnl_port_should_receive(flow)) {
 369         wc->masks.tunnel.tun_id = OVS_BE64_MAX;
 370         if (flow->tunnel.ip_dst) {
 371             wc->masks.tunnel.ip_src = OVS_BE32_MAX;
 372             wc->masks.tunnel.ip_dst = OVS_BE32_MAX;
 373         } else {
 374             wc->masks.tunnel.ipv6_src = in6addr_exact;
 375             wc->masks.tunnel.ipv6_dst = in6addr_exact;
 376         }
 377         wc->masks.tunnel.flags = (FLOW_TNL_F_DONT_FRAGMENT |
 378                                   FLOW_TNL_F_CSUM |
 379                                   FLOW_TNL_F_KEY);
 380         wc->masks.tunnel.ip_tos = UINT8_MAX;
 381         wc->masks.tunnel.ip_ttl = 0;
 382         /* The tp_src and tp_dst members in flow_tnl are set to be always
 383          * wildcarded, not to unwildcard them here. */
 384         wc->masks.tunnel.tp_src = 0;
 385         wc->masks.tunnel.tp_dst = 0;
 386
 387         if (is_ip_any(flow)
 388             && IP_ECN_is_ce(flow->tunnel.ip_tos)) {
 389             wc->masks.nw_tos |= IP_ECN_MASK;
 390         }
 391     }
 392 }
 393
 394 /* Given that 'flow' should be output to the ofport corresponding to
 395  * 'tnl_port', updates 'flow''s tunnel headers and returns the actual datapath
 396  * port that the output should happen on.  May return ODPP_NONE if the output
 397  * shouldn't occur. */
 398 odp_port_t
 399 tnl_port_send(const struct ofport_dpif *ofport, struct flow *flow,
 400               struct flow_wildcards *wc) OVS_EXCLUDED(rwlock)
 401 {
 402     const struct netdev_tunnel_config *cfg;
 403     struct tnl_port *tnl_port;
 404     char *pre_flow_str = NULL;
 405     odp_port_t out_port;
 406
 407     fat_rwlock_rdlock(&rwlock);
 408     tnl_port = tnl_find_ofport(ofport);
 409     out_port = tnl_port ? tnl_port->match.odp_port : ODPP_NONE;
 410     if (!tnl_port) {
 411         goto out;
 412     }
 413
 414     cfg = netdev_get_tunnel_config(tnl_port->netdev);
 415     ovs_assert(cfg);
 416
 417     if (!VLOG_DROP_DBG(&dbg_rl)) {
 418         pre_flow_str = flow_to_string(flow, NULL);
 419     }
 420
 421     if (!cfg->ip_src_flow) {
 422         flow->tunnel.ip_src = in6_addr_get_mapped_ipv4(&tnl_port->match.ipv6_src);
 423         if (!flow->tunnel.ip_src) {
 424             flow->tunnel.ipv6_src = tnl_port->match.ipv6_src;
 425         } else {
 426             flow->tunnel.ipv6_src = in6addr_any;
 427         }
 428     }
 429     if (!cfg->ip_dst_flow) {
 430         flow->tunnel.ip_dst = in6_addr_get_mapped_ipv4(&tnl_port->match.ipv6_dst);
 431         if (!flow->tunnel.ip_dst) {
 432             flow->tunnel.ipv6_dst = tnl_port->match.ipv6_dst;
 433         } else {
 434             flow->tunnel.ipv6_dst = in6addr_any;
 435         }
 436     }
 437     flow->tunnel.tp_dst = cfg->dst_port;
 438     if (!cfg->out_key_flow) {
 439         flow->tunnel.tun_id = cfg->out_key;
 440     }
 441
 442     if (cfg->ttl_inherit && is_ip_any(flow)) {
 443         wc->masks.nw_ttl = 0xff;
 444         flow->tunnel.ip_ttl = flow->nw_ttl;
 445     } else {
 446         flow->tunnel.ip_ttl = cfg->ttl;
 447     }
 448
 449     if (cfg->tos_inherit && is_ip_any(flow)) {
 450         wc->masks.nw_tos |= IP_DSCP_MASK;
 451         flow->tunnel.ip_tos = flow->nw_tos & IP_DSCP_MASK;
 452     } else {
 453         flow->tunnel.ip_tos = cfg->tos;
 454     }
 455
 456     /* ECN fields are always inherited. */
 457     if (is_ip_any(flow)) {
 458         wc->masks.nw_tos |= IP_ECN_MASK;
 459
 460         if (IP_ECN_is_ce(flow->nw_tos)) {
 461             flow->tunnel.ip_tos |= IP_ECN_ECT_0;
 462         } else {
 463             flow->tunnel.ip_tos |= flow->nw_tos & IP_ECN_MASK;
 464         }
 465     }
 466
 467     flow->tunnel.flags &= ~(FLOW_TNL_F_MASK & ~FLOW_TNL_PUB_F_MASK);
 468     flow->tunnel.flags |= (cfg->dont_fragment ? FLOW_TNL_F_DONT_FRAGMENT : 0)
 469         | (cfg->csum ? FLOW_TNL_F_CSUM : 0)
 470         | (cfg->out_key_present ? FLOW_TNL_F_KEY : 0);
 471
 472     if (cfg->set_egress_pkt_mark) {
 473         flow->pkt_mark = cfg->egress_pkt_mark;
 474         wc->masks.pkt_mark = UINT32_MAX;
 475     }
 476
 477     if (!cfg->erspan_ver_flow) {
 478         flow->tunnel.erspan_ver = cfg->erspan_ver;
 479     }
 480
 481     if (!cfg->erspan_idx_flow) {
 482         flow->tunnel.erspan_idx = cfg->erspan_idx;
 483     }
 484
 485     if (!cfg->erspan_dir_flow) {
 486         flow->tunnel.erspan_dir = cfg->erspan_dir;
 487     }
 488
 489     if (!cfg->erspan_hwid_flow) {
 490         flow->tunnel.erspan_hwid = cfg->erspan_hwid;
 491     }
 492
 493     if (pre_flow_str) {
 494         char *post_flow_str = flow_to_string(flow, NULL);
 495         char *tnl_str = tnl_port_to_string(tnl_port);
 496         VLOG_DBG("flow sent\n"
 497                  "%s"
 498                  " pre: %s\n"
 499                  "post: %s",
 500                  tnl_str, pre_flow_str, post_flow_str);
 501         free(tnl_str);
 502         free(pre_flow_str);
 503         free(post_flow_str);
 504     }
 505
 506 out:
 507     fat_rwlock_unlock(&rwlock);
 508     return out_port;
 509 }
 510
 511 static uint32_t
 512 tnl_hash(struct tnl_match *match)
 513 {
 514     BUILD_ASSERT_DECL(sizeof *match % sizeof(uint32_t) == 0);
 515     return hash_words((uint32_t *) match, sizeof *match / sizeof(uint32_t), 0);
 516 }
 517
 518 static struct tnl_port *
 519 tnl_find_ofport(const struct ofport_dpif *ofport) OVS_REQ_RDLOCK(rwlock)
 520 {
 521     struct tnl_port *tnl_port;
 522
 523     HMAP_FOR_EACH_IN_BUCKET (tnl_port, ofport_node, hash_pointer(ofport, 0),
 524                              ofport_map) {
 525         if (tnl_port->ofport == ofport) {
 526             return tnl_port;
 527         }
 528     }
 529     return NULL;
 530 }
 531
 532 static struct tnl_port *
 533 tnl_find_exact(struct tnl_match *match, struct hmap *map)
 534     OVS_REQ_RDLOCK(rwlock)
 535 {
 536     if (map) {
 537         struct tnl_port *tnl_port;
 538
 539         HMAP_FOR_EACH_WITH_HASH (tnl_port, match_node, tnl_hash(match), map) {
 540             if (!memcmp(match, &tnl_port->match, sizeof *match)) {
 541                 return tnl_port;
 542             }
 543         }
 544     }
 545     return NULL;
 546 }
 547
 548 /* Returns the tnl_port that is the best match for the tunnel data in 'flow',
 549  * or NULL if no tnl_port matches 'flow'. */
 550 static struct tnl_port *
 551 tnl_find(const struct flow *flow) OVS_REQ_RDLOCK(rwlock)
 552 {
 553     enum ip_src_type ip_src;
 554     int in_key_flow;
 555     int ip_dst_flow;
 556     int i;
 557
 558     i = 0;
 559     for (in_key_flow = 0; in_key_flow < 2; in_key_flow++) {
 560         for (ip_dst_flow = 0; ip_dst_flow < 2; ip_dst_flow++) {
 561             for (ip_src = 0; ip_src < 3; ip_src++) {
 562                 struct hmap *map = tnl_match_maps[i];
 563
 564                 if (map) {
 565                     struct tnl_port *tnl_port;
 566                     struct tnl_match match;
 567
 568                     memset(&match, 0, sizeof match);
 569
 570                     /* The apparent mix-up of 'ip_dst' and 'ip_src' below is
 571                      * correct, because "struct tnl_match" is expressed in
 572                      * terms of packets being sent out, but we are using it
 573                      * here as a description of how to treat received
 574                      * packets. */
 575                     match.in_key = in_key_flow ? 0 : flow->tunnel.tun_id;
 576                     if (ip_src == IP_SRC_CFG) {
 577                         match.ipv6_src = flow_tnl_dst(&flow->tunnel);
 578                     }
 579                     if (!ip_dst_flow) {
 580                         match.ipv6_dst = flow_tnl_src(&flow->tunnel);
 581                     }
 582                     match.odp_port = flow->in_port.odp_port;
 583                     match.in_key_flow = in_key_flow;
 584                     match.ip_dst_flow = ip_dst_flow;
 585                     match.ip_src_flow = ip_src == IP_SRC_FLOW;
 586
 587                     /* Look for a legacy L2 or L3 tunnel port first. */
 588                     if (pt_ns(flow->packet_type) == OFPHTN_ETHERTYPE) {
 589                         match.pt_mode = NETDEV_PT_LEGACY_L3;
 590                     } else {
 591                         match.pt_mode = NETDEV_PT_LEGACY_L2;
 592                     }
 593                     tnl_port = tnl_find_exact(&match, map);
 594                     if (tnl_port) {
 595                         return tnl_port;
 596                     }
 597
 598                     /* Then check for a packet type aware port. */
 599                     match.pt_mode = NETDEV_PT_AWARE;
 600                     tnl_port = tnl_find_exact(&match, map);
 601                     if (tnl_port) {
 602                         return tnl_port;
 603                     }
 604                 }
 605
 606                 i++;
 607             }
 608         }
 609     }
 610
 611     return NULL;
 612 }
 613
 614 /* Returns a pointer to the 'tnl_match_maps' element corresponding to 'm''s
 615  * matching criteria. */
 616 static struct hmap **
 617 tnl_match_map(const struct tnl_match *m)
 618 {
 619     enum ip_src_type ip_src;
 620
 621     ip_src = (m->ip_src_flow ? IP_SRC_FLOW
 622               : ipv6_addr_is_set(&m->ipv6_src) ? IP_SRC_CFG
 623               : IP_SRC_ANY);
 624
 625     return &tnl_match_maps[6 * m->in_key_flow + 3 * m->ip_dst_flow + ip_src];
 626 }
 627
 628 static void
 629 tnl_match_fmt(const struct tnl_match *match, struct ds *ds)
 630     OVS_REQ_RDLOCK(rwlock)
 631 {
 632     if (!match->ip_dst_flow) {
 633         ipv6_format_mapped(&match->ipv6_src, ds);
 634         ds_put_cstr(ds, "->");
 635         ipv6_format_mapped(&match->ipv6_dst, ds);
 636     } else if (!match->ip_src_flow) {
 637         ipv6_format_mapped(&match->ipv6_src, ds);
 638         ds_put_cstr(ds, "->flow");
 639     } else {
 640         ds_put_cstr(ds, "flow->flow");
 641     }
 642
 643     if (match->in_key_flow) {
 644         ds_put_cstr(ds, ", key=flow");
 645     } else {
 646         ds_put_format(ds, ", key=%#"PRIx64, ntohll(match->in_key));
 647     }
 648
 649     const char *pt_mode
 650         = (match->pt_mode == NETDEV_PT_LEGACY_L2 ? "legacy_l2"
 651            : match->pt_mode == NETDEV_PT_LEGACY_L3 ? "legacy_l3"
 652            : "ptap");
 653     ds_put_format(ds, ", %s, dp port=%"PRIu32, pt_mode, match->odp_port);
 654 }
 655
 656 static void
 657 tnl_port_mod_log(const struct tnl_port *tnl_port, const char *action)
 658     OVS_REQ_RDLOCK(rwlock)
 659 {
 660     if (VLOG_IS_DBG_ENABLED()) {
 661         struct ds ds = DS_EMPTY_INITIALIZER;
 662
 663         tnl_match_fmt(&tnl_port->match, &ds);
 664         VLOG_INFO("%s tunnel port %s (%s)", action,
 665                   tnl_port_get_name(tnl_port), ds_cstr(&ds));
 666         ds_destroy(&ds);
 667     }
 668 }
 669
 670 static void OVS_REQ_RDLOCK(rwlock)
 671 tnl_port_format(const struct tnl_port *tnl_port, struct ds *ds)
 672 {
 673     const struct netdev_tunnel_config *cfg =
 674         netdev_get_tunnel_config(tnl_port->netdev);
 675
 676     ds_put_format(ds, "port %"PRIu32": %s (%s: ", tnl_port->match.odp_port,
 677                   tnl_port_get_name(tnl_port),
 678                   netdev_get_type(tnl_port->netdev));
 679     tnl_match_fmt(&tnl_port->match, ds);
 680
 681     if (cfg->out_key != cfg->in_key ||
 682         cfg->out_key_present != cfg->in_key_present ||
 683         cfg->out_key_flow != cfg->in_key_flow) {
 684         ds_put_cstr(ds, ", out_key=");
 685         if (!cfg->out_key_present) {
 686             ds_put_cstr(ds, "none");
 687         } else if (cfg->out_key_flow) {
 688             ds_put_cstr(ds, "flow");
 689         } else {
 690             ds_put_format(ds, "%#"PRIx64, ntohll(cfg->out_key));
 691         }
 692     }
 693
 694     if (cfg->ttl_inherit) {
 695         ds_put_cstr(ds, ", ttl=inherit");
 696     } else {
 697         ds_put_format(ds, ", ttl=%"PRIu8, cfg->ttl);
 698     }
 699
 700     if (cfg->tos_inherit) {
 701         ds_put_cstr(ds, ", tos=inherit");
 702     } else if (cfg->tos) {
 703         ds_put_format(ds, ", tos=%#"PRIx8, cfg->tos);
 704     }
 705
 706     if (!cfg->dont_fragment) {
 707         ds_put_cstr(ds, ", df=false");
 708     }
 709
 710     if (cfg->csum) {
 711         ds_put_cstr(ds, ", csum=true");
 712     }
 713
 714     ds_put_cstr(ds, ")\n");
 715 }
 716
 717 static char * OVS_REQ_RDLOCK(rwlock)
 718 tnl_port_to_string(const struct tnl_port *tnl_port)
 719 {
 720     struct ds ds = DS_EMPTY_INITIALIZER;
 721     tnl_port_format(tnl_port, &ds);
 722     return ds_steal_cstr(&ds);
 723 }
 724
 725 static const char *
 726 tnl_port_get_name(const struct tnl_port *tnl_port) OVS_REQ_RDLOCK(rwlock)
 727 {
 728     return netdev_get_name(tnl_port->netdev);
 729 }
 730
 731 const char *
 732 tnl_port_get_type(const struct ofport_dpif *ofport) OVS_REQ_RDLOCK(rwlock)
 733 {
 734     struct tnl_port *tnl_port;
 735
 736     tnl_port = tnl_find_ofport(ofport);
 737     return !tnl_port ? NULL :
 738                        netdev_get_type(tnl_port->netdev);
 739 }
 740
 741 int
 742 tnl_port_build_header(const struct ofport_dpif *ofport,
 743                       struct ovs_action_push_tnl *data,
 744                       const struct netdev_tnl_build_header_params *params)
 745 {
 746     struct tnl_port *tnl_port;
 747     int res;
 748
 749     fat_rwlock_rdlock(&rwlock);
 750     tnl_port = tnl_find_ofport(ofport);
 751     ovs_assert(tnl_port);
 752     res = netdev_build_header(tnl_port->netdev, data, params);
 753     fat_rwlock_unlock(&rwlock);
 754
 755     return res;
 756 }
 757
 758 static void
 759 tnl_unixctl_list(struct unixctl_conn *conn,
 760                  int argc OVS_UNUSED, const char *argv[] OVS_UNUSED,
 761                  void *aux OVS_UNUSED)
 762 {
 763     struct ds reply = DS_EMPTY_INITIALIZER;
 764
 765     fat_rwlock_rdlock(&rwlock);
 766     for (int i = 0; i < N_MATCH_TYPES; i++) {
 767         struct hmap *map = tnl_match_maps[i];
 768         if (map) {
 769             struct tnl_port *tnl_port;
 770             HMAP_FOR_EACH (tnl_port, match_node, map) {
 771                 tnl_port_format(tnl_port, &reply);
 772             }
 773         }
 774     }
 775     fat_rwlock_unlock(&rwlock);
 776
 777     unixctl_command_reply(conn, ds_cstr(&reply));
 778     ds_destroy(&reply);
 779 }