2 * Copyright (c) 2014, 2015 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include "tnl-ports.h"
25 #include "classifier.h"
26 #include "openvswitch/dynamic-string.h"
28 #include "openvswitch/list.h"
30 #include "openvswitch/ofpbuf.h"
31 #include "ovs-thread.h"
33 #include "ovs-thread.h"
37 static struct ovs_mutex mutex = OVS_MUTEX_INITIALIZER;
38 static struct classifier cls; /* Tunnel ports. */
43 struct in6_addr *addr;
47 char dev_name[IFNAMSIZ];
50 static struct ovs_list addr_list;
56 char dev_name[IFNAMSIZ];
60 static struct ovs_list port_list;
65 struct ovs_refcount ref_cnt;
66 char dev_name[IFNAMSIZ];
69 static struct tnl_port_in *
70 tnl_port_cast(const struct cls_rule *cr)
72 BUILD_ASSERT_DECL(offsetof(struct tnl_port_in, cr) == 0);
74 return CONTAINER_OF(cr, struct tnl_port_in, cr);
78 tnl_port_free(struct tnl_port_in *p)
80 cls_rule_destroy(&p->cr);
85 tnl_port_init_flow(struct flow *flow, struct eth_addr mac,
86 struct in6_addr *addr, uint8_t nw_proto, ovs_be16 tp_port)
88 memset(flow, 0, sizeof *flow);
91 if (IN6_IS_ADDR_V4MAPPED(addr)) {
92 flow->dl_type = htons(ETH_TYPE_IP);
93 flow->nw_dst = in6_addr_get_mapped_ipv4(addr);
95 flow->dl_type = htons(ETH_TYPE_IPV6);
96 flow->ipv6_dst = *addr;
99 flow->nw_proto = nw_proto;
100 flow->tp_dst = tp_port;
104 map_insert(odp_port_t port, struct eth_addr mac, struct in6_addr *addr,
105 uint8_t nw_proto, ovs_be16 tp_port, const char dev_name[])
107 const struct cls_rule *cr;
108 struct tnl_port_in *p;
111 memset(&match, 0, sizeof match);
112 tnl_port_init_flow(&match.flow, mac, addr, nw_proto, tp_port);
115 cr = classifier_lookup(&cls, CLS_MAX_VERSION, &match.flow, NULL);
116 p = tnl_port_cast(cr);
117 /* Try again if the rule was released before we get the reference. */
118 } while (p && !ovs_refcount_try_ref_rcu(&p->ref_cnt));
121 p = xzalloc(sizeof *p);
124 match.wc.masks.dl_type = OVS_BE16_MAX;
125 match.wc.masks.nw_proto = 0xff;
126 /* XXX: No fragments support. */
127 match.wc.masks.nw_frag = FLOW_NW_FRAG_MASK;
129 /* 'tp_port' is zero for GRE tunnels. In this case it
130 * doesn't make sense to match on UDP port numbers. */
132 match.wc.masks.tp_dst = OVS_BE16_MAX;
134 if (IN6_IS_ADDR_V4MAPPED(addr)) {
135 match.wc.masks.nw_dst = OVS_BE32_MAX;
137 match.wc.masks.ipv6_dst = in6addr_exact;
139 match.wc.masks.vlan_tci = OVS_BE16_MAX;
140 memset(&match.wc.masks.dl_dst, 0xff, sizeof (struct eth_addr));
142 cls_rule_init(&p->cr, &match, 0); /* Priority == 0. */
143 ovs_refcount_init(&p->ref_cnt);
144 ovs_strlcpy(p->dev_name, dev_name, sizeof p->dev_name);
146 classifier_insert(&cls, &p->cr, CLS_MIN_VERSION, NULL, 0);
151 map_insert_ipdev__(struct ip_device *ip_dev, char dev_name[],
152 odp_port_t port, uint8_t nw_proto, ovs_be16 tp_port)
154 if (ip_dev->n_addr) {
157 for (i = 0; i < ip_dev->n_addr; i++) {
158 map_insert(port, ip_dev->mac, &ip_dev->addr[i],
159 nw_proto, tp_port, dev_name);
165 tnl_type_to_nw_proto(const char type[])
167 if (!strcmp(type, "geneve")) {
170 if (!strcmp(type, "stt")) {
173 if (!strcmp(type, "gre")) {
176 if (!strcmp(type, "vxlan")) {
183 tnl_port_map_insert(odp_port_t port, ovs_be16 tp_port,
184 const char dev_name[], const char type[])
187 struct ip_device *ip_dev;
190 nw_proto = tnl_type_to_nw_proto(type);
195 ovs_mutex_lock(&mutex);
196 LIST_FOR_EACH(p, node, &port_list) {
197 if (tp_port == p->tp_port && p->nw_proto == nw_proto) {
202 p = xzalloc(sizeof *p);
204 p->tp_port = tp_port;
205 p->nw_proto = nw_proto;
206 ovs_strlcpy(p->dev_name, dev_name, sizeof p->dev_name);
207 ovs_list_insert(&port_list, &p->node);
209 LIST_FOR_EACH(ip_dev, node, &addr_list) {
210 map_insert_ipdev__(ip_dev, p->dev_name, p->port, p->nw_proto, p->tp_port);
214 ovs_mutex_unlock(&mutex);
218 tnl_port_unref(const struct cls_rule *cr)
220 struct tnl_port_in *p = tnl_port_cast(cr);
222 if (cr && ovs_refcount_unref_relaxed(&p->ref_cnt) == 1) {
223 if (classifier_remove(&cls, cr)) {
224 ovsrcu_postpone(tnl_port_free, p);
230 map_delete(struct eth_addr mac, struct in6_addr *addr,
231 ovs_be16 tp_port, uint8_t nw_proto)
233 const struct cls_rule *cr;
236 tnl_port_init_flow(&flow, mac, addr, nw_proto, tp_port);
238 cr = classifier_lookup(&cls, CLS_MAX_VERSION, &flow, NULL);
243 ipdev_map_delete(struct ip_device *ip_dev, ovs_be16 tp_port, uint8_t nw_proto)
245 if (ip_dev->n_addr) {
248 for (i = 0; i < ip_dev->n_addr; i++) {
249 map_delete(ip_dev->mac, &ip_dev->addr[i], tp_port, nw_proto);
255 tnl_port_map_delete(ovs_be16 tp_port, const char type[])
257 struct tnl_port *p, *next;
258 struct ip_device *ip_dev;
262 nw_proto = tnl_type_to_nw_proto(type);
264 ovs_mutex_lock(&mutex);
265 LIST_FOR_EACH_SAFE(p, next, node, &port_list) {
266 if (p->tp_port == tp_port && p->nw_proto == nw_proto) {
267 ovs_list_remove(&p->node);
276 LIST_FOR_EACH(ip_dev, node, &addr_list) {
277 ipdev_map_delete(ip_dev, p->tp_port, p->nw_proto);
282 ovs_mutex_unlock(&mutex);
285 /* 'flow' is non-const to allow for temporary modifications during the lookup.
286 * Any changes are restored before returning. */
288 tnl_port_map_lookup(struct flow *flow, struct flow_wildcards *wc)
290 const struct cls_rule *cr = classifier_lookup(&cls, CLS_MAX_VERSION, flow,
293 return (cr) ? tnl_port_cast(cr)->portno : ODPP_NONE;
297 tnl_port_show_v(struct ds *ds)
299 const struct tnl_port_in *p;
301 CLS_FOR_EACH(p, cr, &cls) {
302 struct odputil_keybuf keybuf;
303 struct odputil_keybuf maskbuf;
305 const struct nlattr *key, *mask;
306 size_t key_len, mask_len;
307 struct flow_wildcards wc;
309 struct odp_flow_key_parms odp_parms = {
314 ds_put_format(ds, "%s (%"PRIu32") : ", p->dev_name, p->portno);
315 minimask_expand(p->cr.match.mask, &wc);
316 miniflow_expand(p->cr.match.flow, &flow);
319 odp_parms.odp_in_port = flow.in_port.odp_port;
320 odp_parms.support.recirc = true;
321 ofpbuf_use_stack(&buf, &keybuf, sizeof keybuf);
322 odp_flow_key_from_flow(&odp_parms, &buf);
327 odp_parms.odp_in_port = wc.masks.in_port.odp_port;
328 odp_parms.support.recirc = false;
329 ofpbuf_use_stack(&buf, &maskbuf, sizeof maskbuf);
330 odp_flow_key_from_mask(&odp_parms, &buf);
335 odp_flow_format(key, key_len, mask, mask_len, NULL, ds, false);
336 ds_put_format(ds, "\n");
341 tnl_port_show(struct unixctl_conn *conn, int argc OVS_UNUSED,
342 const char *argv[] OVS_UNUSED, void *aux OVS_UNUSED)
344 struct ds ds = DS_EMPTY_INITIALIZER;
347 ds_put_format(&ds, "Listening ports:\n");
348 ovs_mutex_lock(&mutex);
350 if (!strcasecmp(argv[1], "-v")) {
351 tnl_port_show_v(&ds);
356 LIST_FOR_EACH(p, node, &port_list) {
357 ds_put_format(&ds, "%s (%"PRIu32")\n", p->dev_name, p->port);
361 ovs_mutex_unlock(&mutex);
362 unixctl_command_reply(conn, ds_cstr(&ds));
367 map_insert_ipdev(struct ip_device *ip_dev)
371 LIST_FOR_EACH(p, node, &port_list) {
372 map_insert_ipdev__(ip_dev, p->dev_name, p->port, p->nw_proto, p->tp_port);
377 insert_ipdev__(struct netdev *dev,
378 struct in6_addr *addr, int n_addr)
380 struct ip_device *ip_dev;
381 enum netdev_flags flags;
384 error = netdev_get_flags(dev, &flags);
385 if (error || (flags & NETDEV_LOOPBACK)) {
389 ip_dev = xzalloc(sizeof *ip_dev);
390 ip_dev->dev = netdev_ref(dev);
391 ip_dev->change_seq = netdev_get_change_seq(dev);
392 error = netdev_get_etheraddr(ip_dev->dev, &ip_dev->mac);
397 ip_dev->n_addr = n_addr;
398 ovs_strlcpy(ip_dev->dev_name, netdev_get_name(dev), sizeof ip_dev->dev_name);
399 ovs_list_insert(&addr_list, &ip_dev->node);
400 map_insert_ipdev(ip_dev);
404 netdev_close(ip_dev->dev);
411 insert_ipdev(const char dev_name[])
413 struct in6_addr *addr, *mask;
417 error = netdev_open(dev_name, NULL, &dev);
422 error = netdev_get_addr_list(dev, &addr, &mask, &n_in6);
428 insert_ipdev__(dev, addr, n_in6);
433 delete_ipdev(struct ip_device *ip_dev)
437 LIST_FOR_EACH(p, node, &port_list) {
438 ipdev_map_delete(ip_dev, p->tp_port, p->nw_proto);
441 ovs_list_remove(&ip_dev->node);
442 netdev_close(ip_dev->dev);
448 tnl_port_map_insert_ipdev(const char dev_name[])
450 struct ip_device *ip_dev, *next;
452 ovs_mutex_lock(&mutex);
454 LIST_FOR_EACH_SAFE(ip_dev, next, node, &addr_list) {
455 if (!strcmp(netdev_get_name(ip_dev->dev), dev_name)) {
456 if (ip_dev->change_seq == netdev_get_change_seq(ip_dev->dev)) {
459 /* Address changed. */
460 delete_ipdev(ip_dev);
463 insert_ipdev(dev_name);
466 ovs_mutex_unlock(&mutex);
470 tnl_port_map_delete_ipdev(const char dev_name[])
472 struct ip_device *ip_dev, *next;
474 ovs_mutex_lock(&mutex);
475 LIST_FOR_EACH_SAFE(ip_dev, next, node, &addr_list) {
476 if (!strcmp(netdev_get_name(ip_dev->dev), dev_name)) {
477 delete_ipdev(ip_dev);
480 ovs_mutex_unlock(&mutex);
484 tnl_port_map_run(void)
486 struct ip_device *ip_dev, *next;
488 ovs_mutex_lock(&mutex);
489 LIST_FOR_EACH_SAFE(ip_dev, next, node, &addr_list) {
490 char dev_name[IFNAMSIZ];
492 if (ip_dev->change_seq == netdev_get_change_seq(ip_dev->dev)) {
496 /* Address changed. */
497 ovs_strlcpy(dev_name, ip_dev->dev_name, sizeof dev_name);
498 delete_ipdev(ip_dev);
499 insert_ipdev(dev_name);
501 ovs_mutex_unlock(&mutex);
505 tnl_port_map_init(void)
507 classifier_init(&cls, flow_segment_u64s);
508 ovs_list_init(&addr_list);
509 ovs_list_init(&port_list);
510 unixctl_command_register("tnl/ports/show", "-v", 0, 1, tnl_port_show, NULL);