1 /* Copyright (c) 2013 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 "dynamic-string.h"
28 #include "socket-util.h"
32 VLOG_DEFINE_THIS_MODULE(tunnel);
34 /* skb mark used for IPsec tunnel packets */
49 struct hmap_node ofport_node;
50 struct hmap_node match_node;
52 const struct ofport_dpif *ofport;
53 unsigned int netdev_seq;
54 struct netdev *netdev;
56 struct tnl_match match;
59 static struct ovs_rwlock rwlock = OVS_RWLOCK_INITIALIZER;
61 static struct hmap tnl_match_map__ = HMAP_INITIALIZER(&tnl_match_map__);
62 static struct hmap *tnl_match_map OVS_GUARDED_BY(rwlock) = &tnl_match_map__;
64 static struct hmap ofport_map__ = HMAP_INITIALIZER(&ofport_map__);
65 static struct hmap *ofport_map OVS_GUARDED_BY(rwlock) = &ofport_map__;
67 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
68 static struct vlog_rate_limit dbg_rl = VLOG_RATE_LIMIT_INIT(60, 60);
70 static struct tnl_port *tnl_find(const struct flow *) OVS_REQ_RDLOCK(rwlock);
71 static struct tnl_port *tnl_find_exact(struct tnl_match *)
72 OVS_REQ_RDLOCK(rwlock);
73 static struct tnl_port *tnl_find_ofport(const struct ofport_dpif *)
74 OVS_REQ_RDLOCK(rwlock);
76 static uint32_t tnl_hash(struct tnl_match *);
77 static void tnl_match_fmt(const struct tnl_match *, struct ds *);
78 static char *tnl_port_fmt(const struct tnl_port *) OVS_REQ_RDLOCK(rwlock);
79 static void tnl_port_mod_log(const struct tnl_port *, const char *action)
80 OVS_REQ_RDLOCK(rwlock);
81 static const char *tnl_port_get_name(const struct tnl_port *)
82 OVS_REQ_RDLOCK(rwlock);
83 static void tnl_port_del__(const struct ofport_dpif *) OVS_REQ_WRLOCK(rwlock);
86 tnl_port_add__(const struct ofport_dpif *ofport, const struct netdev *netdev,
87 odp_port_t odp_port, bool warn)
88 OVS_REQ_WRLOCK(rwlock)
90 const struct netdev_tunnel_config *cfg;
91 struct tnl_port *existing_port;
92 struct tnl_port *tnl_port;
94 cfg = netdev_get_tunnel_config(netdev);
97 tnl_port = xzalloc(sizeof *tnl_port);
98 tnl_port->ofport = ofport;
99 tnl_port->netdev = netdev_ref(netdev);
100 tnl_port->netdev_seq = netdev_change_seq(tnl_port->netdev);
102 tnl_port->match.in_key = cfg->in_key;
103 tnl_port->match.ip_src = cfg->ip_src;
104 tnl_port->match.ip_dst = cfg->ip_dst;
105 tnl_port->match.ip_src_flow = cfg->ip_src_flow;
106 tnl_port->match.ip_dst_flow = cfg->ip_dst_flow;
107 tnl_port->match.pkt_mark = cfg->ipsec ? IPSEC_MARK : 0;
108 tnl_port->match.in_key_flow = cfg->in_key_flow;
109 tnl_port->match.odp_port = odp_port;
111 existing_port = tnl_find_exact(&tnl_port->match);
114 struct ds ds = DS_EMPTY_INITIALIZER;
115 tnl_match_fmt(&tnl_port->match, &ds);
116 VLOG_WARN("%s: attempting to add tunnel port with same config as "
117 "port '%s' (%s)", tnl_port_get_name(tnl_port),
118 tnl_port_get_name(existing_port), ds_cstr(&ds));
125 hmap_insert(ofport_map, &tnl_port->ofport_node, hash_pointer(ofport, 0));
126 hmap_insert(tnl_match_map, &tnl_port->match_node,
127 tnl_hash(&tnl_port->match));
128 tnl_port_mod_log(tnl_port, "adding");
132 /* Adds 'ofport' to the module with datapath port number 'odp_port'. 'ofport's
133 * must be added before they can be used by the module. 'ofport' must be a
136 tnl_port_add(const struct ofport_dpif *ofport, const struct netdev *netdev,
137 odp_port_t odp_port) OVS_EXCLUDED(rwlock)
139 ovs_rwlock_wrlock(&rwlock);
140 tnl_port_add__(ofport, netdev, odp_port, true);
141 ovs_rwlock_unlock(&rwlock);
144 /* Checks if the tunnel represented by 'ofport' reconfiguration due to changes
145 * in its netdev_tunnel_config. If it does, returns true. Otherwise, returns
146 * false. 'ofport' and 'odp_port' should be the same as would be passed to
149 tnl_port_reconfigure(const struct ofport_dpif *ofport,
150 const struct netdev *netdev, odp_port_t odp_port)
153 struct tnl_port *tnl_port;
154 bool changed = false;
156 ovs_rwlock_wrlock(&rwlock);
157 tnl_port = tnl_find_ofport(ofport);
159 changed = tnl_port_add__(ofport, netdev, odp_port, false);
160 } else if (tnl_port->netdev != netdev
161 || tnl_port->match.odp_port != odp_port
162 || tnl_port->netdev_seq != netdev_change_seq(netdev)) {
163 VLOG_DBG("reconfiguring %s", tnl_port_get_name(tnl_port));
164 tnl_port_del__(ofport);
165 tnl_port_add__(ofport, netdev, odp_port, true);
168 ovs_rwlock_unlock(&rwlock);
173 tnl_port_del__(const struct ofport_dpif *ofport) OVS_REQ_WRLOCK(rwlock)
175 struct tnl_port *tnl_port;
181 tnl_port = tnl_find_ofport(ofport);
183 tnl_port_mod_log(tnl_port, "removing");
184 hmap_remove(tnl_match_map, &tnl_port->match_node);
185 hmap_remove(ofport_map, &tnl_port->ofport_node);
186 netdev_close(tnl_port->netdev);
191 /* Removes 'ofport' from the module. */
193 tnl_port_del(const struct ofport_dpif *ofport) OVS_EXCLUDED(rwlock)
195 ovs_rwlock_wrlock(&rwlock);
196 tnl_port_del__(ofport);
197 ovs_rwlock_unlock(&rwlock);
200 /* Looks in the table of tunnels for a tunnel matching the metadata in 'flow'.
201 * Returns the 'ofport' corresponding to the new in_port, or a null pointer if
204 * Callers should verify that 'flow' needs to be received by calling
205 * tnl_port_should_receive() before this function. */
206 const struct ofport_dpif *
207 tnl_port_receive(const struct flow *flow) OVS_EXCLUDED(rwlock)
209 char *pre_flow_str = NULL;
210 const struct ofport_dpif *ofport;
211 struct tnl_port *tnl_port;
213 ovs_rwlock_rdlock(&rwlock);
214 tnl_port = tnl_find(flow);
215 ofport = tnl_port ? tnl_port->ofport : NULL;
217 char *flow_str = flow_to_string(flow);
219 VLOG_WARN_RL(&rl, "receive tunnel port not found (%s)", flow_str);
224 if (!VLOG_DROP_DBG(&dbg_rl)) {
225 pre_flow_str = flow_to_string(flow);
229 char *post_flow_str = flow_to_string(flow);
230 char *tnl_str = tnl_port_fmt(tnl_port);
231 VLOG_DBG("flow received\n"
235 tnl_str, pre_flow_str, post_flow_str);
242 ovs_rwlock_unlock(&rwlock);
247 tnl_ecn_ok(const struct flow *base_flow, struct flow *flow)
249 if (is_ip_any(base_flow)
250 && (flow->tunnel.ip_tos & IP_ECN_MASK) == IP_ECN_CE) {
251 if ((base_flow->nw_tos & IP_ECN_MASK) == IP_ECN_NOT_ECT) {
252 VLOG_WARN_RL(&rl, "dropping tunnel packet marked ECN CE"
253 " but is not ECN capable");
256 /* Set the ECN CE value in the tunneled packet. */
257 flow->nw_tos |= IP_ECN_CE;
264 /* Should be called at the beginning of action translation to initialize
265 * wildcards and perform any actions based on receiving on tunnel port.
267 * Returns false if the packet must be dropped. */
269 tnl_xlate_init(const struct flow *base_flow, struct flow *flow,
270 struct flow_wildcards *wc)
272 if (tnl_port_should_receive(flow)) {
273 memset(&wc->masks.tunnel, 0xff, sizeof wc->masks.tunnel);
274 memset(&wc->masks.pkt_mark, 0xff, sizeof wc->masks.pkt_mark);
276 if (!tnl_ecn_ok(base_flow, flow)) {
280 flow->pkt_mark &= ~IPSEC_MARK;
286 /* Given that 'flow' should be output to the ofport corresponding to
287 * 'tnl_port', updates 'flow''s tunnel headers and returns the actual datapath
288 * port that the output should happen on. May return ODPP_NONE if the output
289 * shouldn't occur. */
291 tnl_port_send(const struct ofport_dpif *ofport, struct flow *flow,
292 struct flow_wildcards *wc) OVS_EXCLUDED(rwlock)
294 const struct netdev_tunnel_config *cfg;
295 struct tnl_port *tnl_port;
296 char *pre_flow_str = NULL;
299 ovs_rwlock_rdlock(&rwlock);
300 tnl_port = tnl_find_ofport(ofport);
301 out_port = tnl_port ? tnl_port->match.odp_port : ODPP_NONE;
306 cfg = netdev_get_tunnel_config(tnl_port->netdev);
309 if (!VLOG_DROP_DBG(&dbg_rl)) {
310 pre_flow_str = flow_to_string(flow);
313 if (!cfg->ip_src_flow) {
314 flow->tunnel.ip_src = tnl_port->match.ip_src;
316 if (!cfg->ip_dst_flow) {
317 flow->tunnel.ip_dst = tnl_port->match.ip_dst;
319 flow->pkt_mark = tnl_port->match.pkt_mark;
321 if (!cfg->out_key_flow) {
322 flow->tunnel.tun_id = cfg->out_key;
325 if (cfg->ttl_inherit && is_ip_any(flow)) {
326 wc->masks.nw_ttl = 0xff;
327 flow->tunnel.ip_ttl = flow->nw_ttl;
329 flow->tunnel.ip_ttl = cfg->ttl;
332 if (cfg->tos_inherit && is_ip_any(flow)) {
333 wc->masks.nw_tos = 0xff;
334 flow->tunnel.ip_tos = flow->nw_tos & IP_DSCP_MASK;
336 flow->tunnel.ip_tos = cfg->tos;
339 /* ECN fields are always inherited. */
340 if (is_ip_any(flow)) {
341 wc->masks.nw_tos |= IP_ECN_MASK;
344 if ((flow->nw_tos & IP_ECN_MASK) == IP_ECN_CE) {
345 flow->tunnel.ip_tos |= IP_ECN_ECT_0;
347 flow->tunnel.ip_tos |= flow->nw_tos & IP_ECN_MASK;
350 flow->tunnel.flags = (cfg->dont_fragment ? FLOW_TNL_F_DONT_FRAGMENT : 0)
351 | (cfg->csum ? FLOW_TNL_F_CSUM : 0)
352 | (cfg->out_key_present ? FLOW_TNL_F_KEY : 0);
355 char *post_flow_str = flow_to_string(flow);
356 char *tnl_str = tnl_port_fmt(tnl_port);
357 VLOG_DBG("flow sent\n"
361 tnl_str, pre_flow_str, post_flow_str);
368 ovs_rwlock_unlock(&rwlock);
373 tnl_hash(struct tnl_match *match)
375 BUILD_ASSERT_DECL(sizeof *match % sizeof(uint32_t) == 0);
376 return hash_words((uint32_t *) match, sizeof *match / sizeof(uint32_t), 0);
379 static struct tnl_port *
380 tnl_find_ofport(const struct ofport_dpif *ofport) OVS_REQ_RDLOCK(rwlock)
382 struct tnl_port *tnl_port;
384 HMAP_FOR_EACH_IN_BUCKET (tnl_port, ofport_node, hash_pointer(ofport, 0),
386 if (tnl_port->ofport == ofport) {
393 static struct tnl_port *
394 tnl_find_exact(struct tnl_match *match) OVS_REQ_RDLOCK(rwlock)
396 struct tnl_port *tnl_port;
398 HMAP_FOR_EACH_WITH_HASH (tnl_port, match_node, tnl_hash(match),
400 if (!memcmp(match, &tnl_port->match, sizeof *match)) {
407 /* Returns the tnl_port that is the best match for the tunnel data in 'flow',
408 * or NULL if no tnl_port matches 'flow'. */
409 static struct tnl_port *
410 tnl_find(const struct flow *flow) OVS_REQ_RDLOCK(rwlock)
413 IP_SRC_CFG, /* ip_src must equal configured address. */
414 IP_SRC_ANY, /* Any ip_src is acceptable. */
415 IP_SRC_FLOW /* ip_src is handled in flow table. */
418 struct tnl_match_pattern {
421 enum ip_src_type ip_src;
424 static const struct tnl_match_pattern patterns[] = {
425 { false, false, IP_SRC_CFG }, /* remote_ip, local_ip, in_key. */
426 { false, false, IP_SRC_ANY }, /* remote_ip, in_key. */
427 { true, false, IP_SRC_CFG }, /* remote_ip, local_ip. */
428 { true, false, IP_SRC_ANY }, /* remote_ip. */
429 { true, true, IP_SRC_ANY }, /* Flow-based remote. */
430 { true, true, IP_SRC_FLOW }, /* Flow-based everything. */
433 const struct tnl_match_pattern *p;
434 struct tnl_match match;
436 memset(&match, 0, sizeof match);
437 match.odp_port = flow->in_port.odp_port;
438 match.pkt_mark = flow->pkt_mark;
440 for (p = patterns; p < &patterns[ARRAY_SIZE(patterns)]; p++) {
441 struct tnl_port *tnl_port;
443 match.in_key_flow = p->in_key_flow;
444 match.in_key = p->in_key_flow ? 0 : flow->tunnel.tun_id;
446 match.ip_dst_flow = p->ip_dst_flow;
447 match.ip_dst = p->ip_dst_flow ? 0 : flow->tunnel.ip_src;
449 match.ip_src_flow = p->ip_src == IP_SRC_FLOW;
450 match.ip_src = p->ip_src == IP_SRC_CFG ? flow->tunnel.ip_dst : 0;
452 tnl_port = tnl_find_exact(&match);
462 tnl_match_fmt(const struct tnl_match *match, struct ds *ds)
463 OVS_REQ_RDLOCK(rwlock)
465 if (!match->ip_dst_flow) {
466 ds_put_format(ds, IP_FMT"->"IP_FMT, IP_ARGS(match->ip_src),
467 IP_ARGS(match->ip_dst));
468 } else if (!match->ip_src_flow) {
469 ds_put_format(ds, IP_FMT"->flow", IP_ARGS(match->ip_src));
471 ds_put_cstr(ds, "flow->flow");
474 if (match->in_key_flow) {
475 ds_put_cstr(ds, ", key=flow");
477 ds_put_format(ds, ", key=%#"PRIx64, ntohll(match->in_key));
480 ds_put_format(ds, ", dp port=%"PRIu32, match->odp_port);
481 ds_put_format(ds, ", pkt mark=%"PRIu32, match->pkt_mark);
485 tnl_port_mod_log(const struct tnl_port *tnl_port, const char *action)
486 OVS_REQ_RDLOCK(rwlock)
488 if (VLOG_IS_DBG_ENABLED()) {
489 struct ds ds = DS_EMPTY_INITIALIZER;
491 tnl_match_fmt(&tnl_port->match, &ds);
492 VLOG_INFO("%s tunnel port %s (%s)", action,
493 tnl_port_get_name(tnl_port), ds_cstr(&ds));
499 tnl_port_fmt(const struct tnl_port *tnl_port) OVS_REQ_RDLOCK(rwlock)
501 const struct netdev_tunnel_config *cfg =
502 netdev_get_tunnel_config(tnl_port->netdev);
503 struct ds ds = DS_EMPTY_INITIALIZER;
505 ds_put_format(&ds, "port %"PRIu32": %s (%s: ", tnl_port->match.odp_port,
506 tnl_port_get_name(tnl_port),
507 netdev_get_type(tnl_port->netdev));
508 tnl_match_fmt(&tnl_port->match, &ds);
510 if (cfg->out_key != cfg->in_key ||
511 cfg->out_key_present != cfg->in_key_present ||
512 cfg->out_key_flow != cfg->in_key_flow) {
513 ds_put_cstr(&ds, ", out_key=");
514 if (!cfg->out_key_present) {
515 ds_put_cstr(&ds, "none");
516 } else if (cfg->out_key_flow) {
517 ds_put_cstr(&ds, "flow");
519 ds_put_format(&ds, "%#"PRIx64, ntohll(cfg->out_key));
523 if (cfg->ttl_inherit) {
524 ds_put_cstr(&ds, ", ttl=inherit");
526 ds_put_format(&ds, ", ttl=%"PRIu8, cfg->ttl);
529 if (cfg->tos_inherit) {
530 ds_put_cstr(&ds, ", tos=inherit");
531 } else if (cfg->tos) {
532 ds_put_format(&ds, ", tos=%#"PRIx8, cfg->tos);
535 if (!cfg->dont_fragment) {
536 ds_put_cstr(&ds, ", df=false");
540 ds_put_cstr(&ds, ", csum=true");
543 ds_put_cstr(&ds, ")\n");
545 return ds_steal_cstr(&ds);
549 tnl_port_get_name(const struct tnl_port *tnl_port) OVS_REQ_RDLOCK(rwlock)
551 return netdev_get_name(tnl_port->netdev);