1 /* Copyright (c) 2013, 2014 Nicira, Inc.
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:
7 * http://www.apache.org/licenses/LICENSE-2.0
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. */
20 #include "byte-order.h"
21 #include "connectivity.h"
22 #include "dynamic-string.h"
30 #include "socket-util.h"
34 VLOG_DEFINE_THIS_MODULE(tunnel);
36 /* skb mark used for IPsec tunnel packets */
51 struct hmap_node ofport_node;
52 struct hmap_node match_node;
54 const struct ofport_dpif *ofport;
55 unsigned int change_seq;
56 struct netdev *netdev;
58 struct tnl_match match;
61 static struct ovs_rwlock rwlock = OVS_RWLOCK_INITIALIZER;
65 * This module maps packets received over tunnel protocols to vports. The
66 * tunnel protocol and, for some protocols, tunnel-specific information (e.g.,
67 * for VXLAN, the UDP destination port number) are always use as part of the
68 * mapping. Which other fields are used for the mapping depends on the vports
69 * themselves (the parenthesized notations refer to "struct tnl_match" fields):
71 * - in_key: A vport may match a specific tunnel ID (in_key_flow == false)
72 * or arrange for the tunnel ID to be matched as tunnel.tun_id in the
73 * OpenFlow flow (in_key_flow == true).
75 * - ip_dst: A vport may match a specific destination IP address
76 * (ip_dst_flow == false) or arrange for the destination IP to be matched
77 * as tunnel.ip_dst in the OpenFlow flow (ip_dst_flow == true).
79 * - ip_src: A vport may match a specific IP source address (ip_src_flow ==
80 * false, ip_src != 0), wildcard all source addresses (ip_src_flow ==
81 * false, ip_src == 0), or arrange for the IP source address to be
82 * handled in the OpenFlow flow table (ip_src_flow == true).
84 * Thus, there are 2 * 2 * 3 == 12 possible ways a vport can match against a
85 * tunnel packet. We number the possibilities for each field in increasing
86 * order as listed in each bullet above. We order the 12 overall combinations
87 * in lexicographic order considering in_key first, then ip_dst, then
89 #define N_MATCH_TYPES (2 * 2 * 3)
91 /* The three possibilities (see above) for vport ip_src matches. */
93 IP_SRC_CFG, /* ip_src must equal configured address. */
94 IP_SRC_ANY, /* Any ip_src is acceptable. */
95 IP_SRC_FLOW /* ip_src is handled in flow table. */
98 /* Each hmap contains "struct tnl_port"s.
99 * The index is a combination of how each of the fields listed under "Tunnel
100 * matches" above matches, see the final paragraph for ordering. */
101 static struct hmap *tnl_match_maps[N_MATCH_TYPES] OVS_GUARDED_BY(rwlock);
102 static struct hmap **tnl_match_map(const struct tnl_match *);
104 static struct hmap ofport_map__ = HMAP_INITIALIZER(&ofport_map__);
105 static struct hmap *ofport_map OVS_GUARDED_BY(rwlock) = &ofport_map__;
107 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
108 static struct vlog_rate_limit dbg_rl = VLOG_RATE_LIMIT_INIT(60, 60);
110 static struct tnl_port *tnl_find(const struct flow *) OVS_REQ_RDLOCK(rwlock);
111 static struct tnl_port *tnl_find_exact(struct tnl_match *, struct hmap *)
112 OVS_REQ_RDLOCK(rwlock);
113 static struct tnl_port *tnl_find_ofport(const struct ofport_dpif *)
114 OVS_REQ_RDLOCK(rwlock);
116 static uint32_t tnl_hash(struct tnl_match *);
117 static void tnl_match_fmt(const struct tnl_match *, struct ds *);
118 static char *tnl_port_fmt(const struct tnl_port *) OVS_REQ_RDLOCK(rwlock);
119 static void tnl_port_mod_log(const struct tnl_port *, const char *action)
120 OVS_REQ_RDLOCK(rwlock);
121 static const char *tnl_port_get_name(const struct tnl_port *)
122 OVS_REQ_RDLOCK(rwlock);
123 static void tnl_port_del__(const struct ofport_dpif *) OVS_REQ_WRLOCK(rwlock);
126 tnl_port_add__(const struct ofport_dpif *ofport, const struct netdev *netdev,
127 odp_port_t odp_port, bool warn)
128 OVS_REQ_WRLOCK(rwlock)
130 const struct netdev_tunnel_config *cfg;
131 struct tnl_port *existing_port;
132 struct tnl_port *tnl_port;
135 cfg = netdev_get_tunnel_config(netdev);
138 tnl_port = xzalloc(sizeof *tnl_port);
139 tnl_port->ofport = ofport;
140 tnl_port->netdev = netdev_ref(netdev);
141 tnl_port->change_seq = seq_read(connectivity_seq_get());
143 tnl_port->match.in_key = cfg->in_key;
144 tnl_port->match.ip_src = cfg->ip_src;
145 tnl_port->match.ip_dst = cfg->ip_dst;
146 tnl_port->match.ip_src_flow = cfg->ip_src_flow;
147 tnl_port->match.ip_dst_flow = cfg->ip_dst_flow;
148 tnl_port->match.pkt_mark = cfg->ipsec ? IPSEC_MARK : 0;
149 tnl_port->match.in_key_flow = cfg->in_key_flow;
150 tnl_port->match.odp_port = odp_port;
152 map = tnl_match_map(&tnl_port->match);
153 existing_port = tnl_find_exact(&tnl_port->match, *map);
156 struct ds ds = DS_EMPTY_INITIALIZER;
157 tnl_match_fmt(&tnl_port->match, &ds);
158 VLOG_WARN("%s: attempting to add tunnel port with same config as "
159 "port '%s' (%s)", tnl_port_get_name(tnl_port),
160 tnl_port_get_name(existing_port), ds_cstr(&ds));
167 hmap_insert(ofport_map, &tnl_port->ofport_node, hash_pointer(ofport, 0));
170 *map = xmalloc(sizeof **map);
173 hmap_insert(*map, &tnl_port->match_node, tnl_hash(&tnl_port->match));
174 tnl_port_mod_log(tnl_port, "adding");
178 /* Adds 'ofport' to the module with datapath port number 'odp_port'. 'ofport's
179 * must be added before they can be used by the module. 'ofport' must be a
182 tnl_port_add(const struct ofport_dpif *ofport, const struct netdev *netdev,
183 odp_port_t odp_port) OVS_EXCLUDED(rwlock)
185 ovs_rwlock_wrlock(&rwlock);
186 tnl_port_add__(ofport, netdev, odp_port, true);
187 ovs_rwlock_unlock(&rwlock);
190 /* Checks if the tunnel represented by 'ofport' reconfiguration due to changes
191 * in its netdev_tunnel_config. If it does, returns true. Otherwise, returns
192 * false. 'ofport' and 'odp_port' should be the same as would be passed to
195 tnl_port_reconfigure(const struct ofport_dpif *ofport,
196 const struct netdev *netdev, odp_port_t odp_port)
199 struct tnl_port *tnl_port;
200 bool changed = false;
202 ovs_rwlock_wrlock(&rwlock);
203 tnl_port = tnl_find_ofport(ofport);
205 changed = tnl_port_add__(ofport, netdev, odp_port, false);
206 } else if (tnl_port->netdev != netdev
207 || tnl_port->match.odp_port != odp_port
208 || tnl_port->change_seq != seq_read(connectivity_seq_get())) {
209 VLOG_DBG("reconfiguring %s", tnl_port_get_name(tnl_port));
210 tnl_port_del__(ofport);
211 tnl_port_add__(ofport, netdev, odp_port, true);
214 ovs_rwlock_unlock(&rwlock);
219 tnl_port_del__(const struct ofport_dpif *ofport) OVS_REQ_WRLOCK(rwlock)
221 struct tnl_port *tnl_port;
227 tnl_port = tnl_find_ofport(ofport);
231 tnl_port_mod_log(tnl_port, "removing");
232 map = tnl_match_map(&tnl_port->match);
233 hmap_remove(*map, &tnl_port->match_node);
234 if (hmap_is_empty(*map)) {
239 hmap_remove(ofport_map, &tnl_port->ofport_node);
240 netdev_close(tnl_port->netdev);
245 /* Removes 'ofport' from the module. */
247 tnl_port_del(const struct ofport_dpif *ofport) OVS_EXCLUDED(rwlock)
249 ovs_rwlock_wrlock(&rwlock);
250 tnl_port_del__(ofport);
251 ovs_rwlock_unlock(&rwlock);
254 /* Looks in the table of tunnels for a tunnel matching the metadata in 'flow'.
255 * Returns the 'ofport' corresponding to the new in_port, or a null pointer if
258 * Callers should verify that 'flow' needs to be received by calling
259 * tnl_port_should_receive() before this function. */
260 const struct ofport_dpif *
261 tnl_port_receive(const struct flow *flow) OVS_EXCLUDED(rwlock)
263 char *pre_flow_str = NULL;
264 const struct ofport_dpif *ofport;
265 struct tnl_port *tnl_port;
267 ovs_rwlock_rdlock(&rwlock);
268 tnl_port = tnl_find(flow);
269 ofport = tnl_port ? tnl_port->ofport : NULL;
271 char *flow_str = flow_to_string(flow);
273 VLOG_WARN_RL(&rl, "receive tunnel port not found (%s)", flow_str);
278 if (!VLOG_DROP_DBG(&dbg_rl)) {
279 pre_flow_str = flow_to_string(flow);
283 char *post_flow_str = flow_to_string(flow);
284 char *tnl_str = tnl_port_fmt(tnl_port);
285 VLOG_DBG("flow received\n"
289 tnl_str, pre_flow_str, post_flow_str);
296 ovs_rwlock_unlock(&rwlock);
301 tnl_ecn_ok(const struct flow *base_flow, struct flow *flow,
302 struct flow_wildcards *wc)
304 if (is_ip_any(base_flow)) {
305 wc->masks.nw_tos |= IP_ECN_MASK;
306 if ((flow->tunnel.ip_tos & IP_ECN_MASK) == IP_ECN_CE) {
307 if ((base_flow->nw_tos & IP_ECN_MASK) == IP_ECN_NOT_ECT) {
308 VLOG_WARN_RL(&rl, "dropping tunnel packet marked ECN CE"
309 " but is not ECN capable");
312 /* Set the ECN CE value in the tunneled packet. */
313 flow->nw_tos |= IP_ECN_CE;
321 /* Should be called at the beginning of action translation to initialize
322 * wildcards and perform any actions based on receiving on tunnel port.
324 * Returns false if the packet must be dropped. */
326 tnl_xlate_init(const struct flow *base_flow, struct flow *flow,
327 struct flow_wildcards *wc)
329 if (tnl_port_should_receive(flow)) {
330 wc->masks.tunnel.tun_id = OVS_BE64_MAX;
331 wc->masks.tunnel.ip_src = OVS_BE32_MAX;
332 wc->masks.tunnel.ip_dst = OVS_BE32_MAX;
333 wc->masks.tunnel.flags = (FLOW_TNL_F_DONT_FRAGMENT |
336 wc->masks.tunnel.ip_tos = UINT8_MAX;
337 wc->masks.tunnel.ip_ttl = UINT8_MAX;
339 memset(&wc->masks.pkt_mark, 0xff, sizeof wc->masks.pkt_mark);
341 if (!tnl_ecn_ok(base_flow, flow, wc)) {
345 flow->pkt_mark &= ~IPSEC_MARK;
351 /* Given that 'flow' should be output to the ofport corresponding to
352 * 'tnl_port', updates 'flow''s tunnel headers and returns the actual datapath
353 * port that the output should happen on. May return ODPP_NONE if the output
354 * shouldn't occur. */
356 tnl_port_send(const struct ofport_dpif *ofport, struct flow *flow,
357 struct flow_wildcards *wc) OVS_EXCLUDED(rwlock)
359 const struct netdev_tunnel_config *cfg;
360 struct tnl_port *tnl_port;
361 char *pre_flow_str = NULL;
364 ovs_rwlock_rdlock(&rwlock);
365 tnl_port = tnl_find_ofport(ofport);
366 out_port = tnl_port ? tnl_port->match.odp_port : ODPP_NONE;
371 cfg = netdev_get_tunnel_config(tnl_port->netdev);
374 if (!VLOG_DROP_DBG(&dbg_rl)) {
375 pre_flow_str = flow_to_string(flow);
378 if (!cfg->ip_src_flow) {
379 flow->tunnel.ip_src = tnl_port->match.ip_src;
381 if (!cfg->ip_dst_flow) {
382 flow->tunnel.ip_dst = tnl_port->match.ip_dst;
384 flow->pkt_mark = tnl_port->match.pkt_mark;
386 if (!cfg->out_key_flow) {
387 flow->tunnel.tun_id = cfg->out_key;
390 if (cfg->ttl_inherit && is_ip_any(flow)) {
391 wc->masks.nw_ttl = 0xff;
392 flow->tunnel.ip_ttl = flow->nw_ttl;
394 flow->tunnel.ip_ttl = cfg->ttl;
397 if (cfg->tos_inherit && is_ip_any(flow)) {
398 wc->masks.nw_tos = IP_DSCP_MASK;
399 flow->tunnel.ip_tos = flow->nw_tos & IP_DSCP_MASK;
401 flow->tunnel.ip_tos = cfg->tos;
404 /* ECN fields are always inherited. */
405 if (is_ip_any(flow)) {
406 wc->masks.nw_tos |= IP_ECN_MASK;
408 if ((flow->nw_tos & IP_ECN_MASK) == IP_ECN_CE) {
409 flow->tunnel.ip_tos |= IP_ECN_ECT_0;
411 flow->tunnel.ip_tos |= flow->nw_tos & IP_ECN_MASK;
415 flow->tunnel.flags = (cfg->dont_fragment ? FLOW_TNL_F_DONT_FRAGMENT : 0)
416 | (cfg->csum ? FLOW_TNL_F_CSUM : 0)
417 | (cfg->out_key_present ? FLOW_TNL_F_KEY : 0);
420 char *post_flow_str = flow_to_string(flow);
421 char *tnl_str = tnl_port_fmt(tnl_port);
422 VLOG_DBG("flow sent\n"
426 tnl_str, pre_flow_str, post_flow_str);
433 ovs_rwlock_unlock(&rwlock);
438 tnl_hash(struct tnl_match *match)
440 BUILD_ASSERT_DECL(sizeof *match % sizeof(uint32_t) == 0);
441 return hash_words((uint32_t *) match, sizeof *match / sizeof(uint32_t), 0);
444 static struct tnl_port *
445 tnl_find_ofport(const struct ofport_dpif *ofport) OVS_REQ_RDLOCK(rwlock)
447 struct tnl_port *tnl_port;
449 HMAP_FOR_EACH_IN_BUCKET (tnl_port, ofport_node, hash_pointer(ofport, 0),
451 if (tnl_port->ofport == ofport) {
458 static struct tnl_port *
459 tnl_find_exact(struct tnl_match *match, struct hmap *map)
460 OVS_REQ_RDLOCK(rwlock)
463 struct tnl_port *tnl_port;
465 HMAP_FOR_EACH_WITH_HASH (tnl_port, match_node, tnl_hash(match), map) {
466 if (!memcmp(match, &tnl_port->match, sizeof *match)) {
474 /* Returns the tnl_port that is the best match for the tunnel data in 'flow',
475 * or NULL if no tnl_port matches 'flow'. */
476 static struct tnl_port *
477 tnl_find(const struct flow *flow) OVS_REQ_RDLOCK(rwlock)
479 enum ip_src_type ip_src;
485 for (in_key_flow = 0; in_key_flow < 2; in_key_flow++) {
486 for (ip_dst_flow = 0; ip_dst_flow < 2; ip_dst_flow++) {
487 for (ip_src = 0; ip_src < 3; ip_src++) {
488 struct hmap *map = tnl_match_maps[i];
491 struct tnl_port *tnl_port;
492 struct tnl_match match;
494 memset(&match, 0, sizeof match);
496 /* The apparent mix-up of 'ip_dst' and 'ip_src' below is
497 * correct, because "struct tnl_match" is expressed in
498 * terms of packets being sent out, but we are using it
499 * here as a description of how to treat received
501 match.in_key = in_key_flow ? 0 : flow->tunnel.tun_id;
502 match.ip_src = (ip_src == IP_SRC_CFG
503 ? flow->tunnel.ip_dst
505 match.ip_dst = ip_dst_flow ? 0 : flow->tunnel.ip_src;
506 match.odp_port = flow->in_port.odp_port;
507 match.pkt_mark = flow->pkt_mark;
508 match.in_key_flow = in_key_flow;
509 match.ip_dst_flow = ip_dst_flow;
510 match.ip_src_flow = ip_src == IP_SRC_FLOW;
512 tnl_port = tnl_find_exact(&match, map);
526 /* Returns a pointer to the 'tnl_match_maps' element corresponding to 'm''s
527 * matching criteria. */
528 static struct hmap **
529 tnl_match_map(const struct tnl_match *m)
531 enum ip_src_type ip_src;
533 ip_src = (m->ip_src_flow ? IP_SRC_FLOW
534 : m->ip_src ? IP_SRC_CFG
537 return &tnl_match_maps[6 * m->in_key_flow + 3 * m->ip_dst_flow + ip_src];
541 tnl_match_fmt(const struct tnl_match *match, struct ds *ds)
542 OVS_REQ_RDLOCK(rwlock)
544 if (!match->ip_dst_flow) {
545 ds_put_format(ds, IP_FMT"->"IP_FMT, IP_ARGS(match->ip_src),
546 IP_ARGS(match->ip_dst));
547 } else if (!match->ip_src_flow) {
548 ds_put_format(ds, IP_FMT"->flow", IP_ARGS(match->ip_src));
550 ds_put_cstr(ds, "flow->flow");
553 if (match->in_key_flow) {
554 ds_put_cstr(ds, ", key=flow");
556 ds_put_format(ds, ", key=%#"PRIx64, ntohll(match->in_key));
559 ds_put_format(ds, ", dp port=%"PRIu32, match->odp_port);
560 ds_put_format(ds, ", pkt mark=%"PRIu32, match->pkt_mark);
564 tnl_port_mod_log(const struct tnl_port *tnl_port, const char *action)
565 OVS_REQ_RDLOCK(rwlock)
567 if (VLOG_IS_DBG_ENABLED()) {
568 struct ds ds = DS_EMPTY_INITIALIZER;
570 tnl_match_fmt(&tnl_port->match, &ds);
571 VLOG_INFO("%s tunnel port %s (%s)", action,
572 tnl_port_get_name(tnl_port), ds_cstr(&ds));
578 tnl_port_fmt(const struct tnl_port *tnl_port) OVS_REQ_RDLOCK(rwlock)
580 const struct netdev_tunnel_config *cfg =
581 netdev_get_tunnel_config(tnl_port->netdev);
582 struct ds ds = DS_EMPTY_INITIALIZER;
584 ds_put_format(&ds, "port %"PRIu32": %s (%s: ", tnl_port->match.odp_port,
585 tnl_port_get_name(tnl_port),
586 netdev_get_type(tnl_port->netdev));
587 tnl_match_fmt(&tnl_port->match, &ds);
589 if (cfg->out_key != cfg->in_key ||
590 cfg->out_key_present != cfg->in_key_present ||
591 cfg->out_key_flow != cfg->in_key_flow) {
592 ds_put_cstr(&ds, ", out_key=");
593 if (!cfg->out_key_present) {
594 ds_put_cstr(&ds, "none");
595 } else if (cfg->out_key_flow) {
596 ds_put_cstr(&ds, "flow");
598 ds_put_format(&ds, "%#"PRIx64, ntohll(cfg->out_key));
602 if (cfg->ttl_inherit) {
603 ds_put_cstr(&ds, ", ttl=inherit");
605 ds_put_format(&ds, ", ttl=%"PRIu8, cfg->ttl);
608 if (cfg->tos_inherit) {
609 ds_put_cstr(&ds, ", tos=inherit");
610 } else if (cfg->tos) {
611 ds_put_format(&ds, ", tos=%#"PRIx8, cfg->tos);
614 if (!cfg->dont_fragment) {
615 ds_put_cstr(&ds, ", df=false");
619 ds_put_cstr(&ds, ", csum=true");
622 ds_put_cstr(&ds, ")\n");
624 return ds_steal_cstr(&ds);
628 tnl_port_get_name(const struct tnl_port *tnl_port) OVS_REQ_RDLOCK(rwlock)
630 return netdev_get_name(tnl_port->netdev);