2 * Copyright (c) 2010, 2011, 2012, 2013 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 "netdev-vport.h"
23 #include <sys/socket.h>
25 #include <sys/ioctl.h>
27 #include "byte-order.h"
34 #include "netdev-provider.h"
37 #include "route-table.h"
39 #include "socket-util.h"
42 VLOG_DEFINE_THIS_MODULE(netdev_vport);
44 #define VXLAN_DST_PORT 4789
45 #define LISP_DST_PORT 4341
47 #define DEFAULT_TTL 64
51 unsigned int change_seq;
52 uint8_t etheraddr[ETH_ADDR_LEN];
53 struct netdev_stats stats;
56 struct netdev_tunnel_config tnl_cfg;
63 const char *dpif_port;
64 struct netdev_class netdev_class;
67 static int netdev_vport_create(const struct netdev_class *, const char *,
69 static int get_patch_config(const struct netdev *, struct smap *args);
70 static int get_tunnel_config(const struct netdev *, struct smap *args);
71 static void netdev_vport_poll_notify(struct netdev_vport *);
74 is_vport_class(const struct netdev_class *class)
76 return class->create == netdev_vport_create;
79 static const struct vport_class *
80 vport_class_cast(const struct netdev_class *class)
82 ovs_assert(is_vport_class(class));
83 return CONTAINER_OF(class, struct vport_class, netdev_class);
86 static struct netdev_vport *
87 netdev_vport_cast(const struct netdev *netdev)
89 ovs_assert(is_vport_class(netdev_get_class(netdev)));
90 return CONTAINER_OF(netdev, struct netdev_vport, up);
93 static const struct netdev_tunnel_config *
94 get_netdev_tunnel_config(const struct netdev *netdev)
96 return &netdev_vport_cast(netdev)->tnl_cfg;
100 netdev_vport_is_patch(const struct netdev *netdev)
102 const struct netdev_class *class = netdev_get_class(netdev);
104 return class->get_config == get_patch_config;
108 netdev_vport_needs_dst_port(const struct netdev *dev)
110 const struct netdev_class *class = netdev_get_class(dev);
111 const char *type = netdev_get_type(dev);
113 return (class->get_config == get_tunnel_config &&
114 (!strcmp("vxlan", type) || !strcmp("lisp", type)));
118 netdev_vport_class_get_dpif_port(const struct netdev_class *class)
120 return is_vport_class(class) ? vport_class_cast(class)->dpif_port : NULL;
124 netdev_vport_get_dpif_port(const struct netdev *netdev)
126 const char *dpif_port;
128 if (netdev_vport_needs_dst_port(netdev)) {
129 const struct netdev_vport *vport = netdev_vport_cast(netdev);
130 const char *type = netdev_get_type(netdev);
131 static char dpif_port_combined[IFNAMSIZ];
134 * Note: IFNAMSIZ is 16 bytes long. The maximum length of a VXLAN
135 * or LISP port name below is 15 or 14 bytes respectively. Still,
136 * assert here on the size of strlen(type) in case that changes
139 ovs_assert(strlen(type) + 10 < IFNAMSIZ);
140 snprintf(dpif_port_combined, IFNAMSIZ, "%s_sys_%d", type,
141 ntohs(vport->tnl_cfg.dst_port));
142 return dpif_port_combined;
144 const struct netdev_class *class = netdev_get_class(netdev);
145 dpif_port = netdev_vport_class_get_dpif_port(class);
148 return dpif_port ? dpif_port : netdev_get_name(netdev);
152 netdev_vport_create(const struct netdev_class *netdev_class, const char *name,
153 struct netdev **netdevp)
155 struct netdev_vport *dev;
157 dev = xzalloc(sizeof *dev);
158 netdev_init(&dev->up, name, netdev_class);
160 eth_addr_random(dev->etheraddr);
163 route_table_register();
169 netdev_vport_destroy(struct netdev *netdev_)
171 struct netdev_vport *netdev = netdev_vport_cast(netdev_);
173 route_table_unregister();
179 netdev_vport_set_etheraddr(struct netdev *netdev_,
180 const uint8_t mac[ETH_ADDR_LEN])
182 struct netdev_vport *netdev = netdev_vport_cast(netdev_);
183 memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN);
184 netdev_vport_poll_notify(netdev);
189 netdev_vport_get_etheraddr(const struct netdev *netdev,
190 uint8_t mac[ETH_ADDR_LEN])
192 memcpy(mac, netdev_vport_cast(netdev)->etheraddr, ETH_ADDR_LEN);
197 tunnel_get_status(const struct netdev *netdev, struct smap *smap)
199 static char iface[IFNAMSIZ];
202 route = netdev_vport_cast(netdev)->tnl_cfg.ip_dst;
203 if (route_table_get_name(route, iface)) {
204 struct netdev *egress_netdev;
206 smap_add(smap, "tunnel_egress_iface", iface);
208 if (!netdev_open(iface, "system", &egress_netdev)) {
209 smap_add(smap, "tunnel_egress_iface_carrier",
210 netdev_get_carrier(egress_netdev) ? "up" : "down");
211 netdev_close(egress_netdev);
219 netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED,
220 enum netdev_flags off,
221 enum netdev_flags on OVS_UNUSED,
222 enum netdev_flags *old_flagsp)
224 if (off & (NETDEV_UP | NETDEV_PROMISC)) {
228 *old_flagsp = NETDEV_UP | NETDEV_PROMISC;
233 netdev_vport_change_seq(const struct netdev *netdev)
235 return netdev_vport_cast(netdev)->change_seq;
239 netdev_vport_run(void)
245 netdev_vport_wait(void)
250 /* Helper functions. */
253 netdev_vport_poll_notify(struct netdev_vport *ndv)
256 if (!ndv->change_seq) {
261 /* Code specific to tunnel types. */
264 parse_key(const struct smap *args, const char *name,
265 bool *present, bool *flow)
272 s = smap_get(args, name);
274 s = smap_get(args, "key");
282 if (!strcmp(s, "flow")) {
286 return htonll(strtoull(s, NULL, 0));
291 set_tunnel_config(struct netdev *dev_, const struct smap *args)
293 struct netdev_vport *dev = netdev_vport_cast(dev_);
294 const char *name = netdev_get_name(dev_);
295 const char *type = netdev_get_type(dev_);
296 bool ipsec_mech_set, needs_dst_port, has_csum;
297 struct netdev_tunnel_config tnl_cfg;
298 struct smap_node *node;
300 has_csum = strstr(type, "gre");
301 ipsec_mech_set = false;
302 memset(&tnl_cfg, 0, sizeof tnl_cfg);
304 needs_dst_port = netdev_vport_needs_dst_port(dev_);
305 tnl_cfg.ipsec = strstr(type, "ipsec");
306 tnl_cfg.dont_fragment = true;
308 SMAP_FOR_EACH (node, args) {
309 if (!strcmp(node->key, "remote_ip")) {
310 struct in_addr in_addr;
311 if (!strcmp(node->value, "flow")) {
312 tnl_cfg.ip_dst_flow = true;
313 tnl_cfg.ip_dst = htonl(0);
314 } else if (lookup_ip(node->value, &in_addr)) {
315 VLOG_WARN("%s: bad %s 'remote_ip'", name, type);
316 } else if (ip_is_multicast(in_addr.s_addr)) {
317 VLOG_WARN("%s: multicast remote_ip="IP_FMT" not allowed",
318 name, IP_ARGS(in_addr.s_addr));
321 tnl_cfg.ip_dst = in_addr.s_addr;
323 } else if (!strcmp(node->key, "local_ip")) {
324 struct in_addr in_addr;
325 if (!strcmp(node->value, "flow")) {
326 tnl_cfg.ip_src_flow = true;
327 tnl_cfg.ip_src = htonl(0);
328 } else if (lookup_ip(node->value, &in_addr)) {
329 VLOG_WARN("%s: bad %s 'local_ip'", name, type);
331 tnl_cfg.ip_src = in_addr.s_addr;
333 } else if (!strcmp(node->key, "tos")) {
334 if (!strcmp(node->value, "inherit")) {
335 tnl_cfg.tos_inherit = true;
339 tos = strtol(node->value, &endptr, 0);
340 if (*endptr == '\0' && tos == (tos & IP_DSCP_MASK)) {
343 VLOG_WARN("%s: invalid TOS %s", name, node->value);
346 } else if (!strcmp(node->key, "ttl")) {
347 if (!strcmp(node->value, "inherit")) {
348 tnl_cfg.ttl_inherit = true;
350 tnl_cfg.ttl = atoi(node->value);
352 } else if (!strcmp(node->key, "dst_port") && needs_dst_port) {
353 tnl_cfg.dst_port = htons(atoi(node->value));
354 } else if (!strcmp(node->key, "csum") && has_csum) {
355 if (!strcmp(node->value, "true")) {
358 } else if (!strcmp(node->key, "df_default")) {
359 if (!strcmp(node->value, "false")) {
360 tnl_cfg.dont_fragment = false;
362 } else if (!strcmp(node->key, "peer_cert") && tnl_cfg.ipsec) {
363 if (smap_get(args, "certificate")) {
364 ipsec_mech_set = true;
366 const char *use_ssl_cert;
368 /* If the "use_ssl_cert" is true, then "certificate" and
369 * "private_key" will be pulled from the SSL table. The
370 * use of this option is strongly discouraged, since it
371 * will like be removed when multiple SSL configurations
372 * are supported by OVS.
374 use_ssl_cert = smap_get(args, "use_ssl_cert");
375 if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) {
376 VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument",
380 ipsec_mech_set = true;
382 } else if (!strcmp(node->key, "psk") && tnl_cfg.ipsec) {
383 ipsec_mech_set = true;
384 } else if (tnl_cfg.ipsec
385 && (!strcmp(node->key, "certificate")
386 || !strcmp(node->key, "private_key")
387 || !strcmp(node->key, "use_ssl_cert"))) {
388 /* Ignore options not used by the netdev. */
389 } else if (!strcmp(node->key, "key") ||
390 !strcmp(node->key, "in_key") ||
391 !strcmp(node->key, "out_key")) {
392 /* Handled separately below. */
394 VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key);
398 /* Add a default destination port for VXLAN if none specified. */
399 if (!strcmp(type, "vxlan") && !tnl_cfg.dst_port) {
400 tnl_cfg.dst_port = htons(VXLAN_DST_PORT);
403 /* Add a default destination port for LISP if none specified. */
404 if (!strcmp(type, "lisp") && !tnl_cfg.dst_port) {
405 tnl_cfg.dst_port = htons(LISP_DST_PORT);
409 static pid_t pid = 0;
411 char *file_name = xasprintf("%s/%s", ovs_rundir(),
412 "ovs-monitor-ipsec.pid");
413 pid = read_pidfile(file_name);
418 VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon",
423 if (smap_get(args, "peer_cert") && smap_get(args, "psk")) {
424 VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name);
428 if (!ipsec_mech_set) {
429 VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument",
435 if (!tnl_cfg.ip_dst && !tnl_cfg.ip_dst_flow) {
436 VLOG_ERR("%s: %s type requires valid 'remote_ip' argument",
440 if (tnl_cfg.ip_src_flow && !tnl_cfg.ip_dst_flow) {
441 VLOG_ERR("%s: %s type requires 'remote_ip=flow' with 'local_ip=flow'",
446 tnl_cfg.ttl = DEFAULT_TTL;
449 tnl_cfg.in_key = parse_key(args, "in_key",
450 &tnl_cfg.in_key_present,
451 &tnl_cfg.in_key_flow);
453 tnl_cfg.out_key = parse_key(args, "out_key",
454 &tnl_cfg.out_key_present,
455 &tnl_cfg.out_key_flow);
457 dev->tnl_cfg = tnl_cfg;
458 netdev_vport_poll_notify(dev);
464 get_tunnel_config(const struct netdev *dev, struct smap *args)
466 const struct netdev_tunnel_config *tnl_cfg =
467 &netdev_vport_cast(dev)->tnl_cfg;
469 if (tnl_cfg->ip_dst) {
470 smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(tnl_cfg->ip_dst));
471 } else if (tnl_cfg->ip_dst_flow) {
472 smap_add(args, "remote_ip", "flow");
475 if (tnl_cfg->ip_src) {
476 smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(tnl_cfg->ip_src));
477 } else if (tnl_cfg->ip_src_flow) {
478 smap_add(args, "local_ip", "flow");
481 if (tnl_cfg->in_key_flow && tnl_cfg->out_key_flow) {
482 smap_add(args, "key", "flow");
483 } else if (tnl_cfg->in_key_present && tnl_cfg->out_key_present
484 && tnl_cfg->in_key == tnl_cfg->out_key) {
485 smap_add_format(args, "key", "%"PRIu64, ntohll(tnl_cfg->in_key));
487 if (tnl_cfg->in_key_flow) {
488 smap_add(args, "in_key", "flow");
489 } else if (tnl_cfg->in_key_present) {
490 smap_add_format(args, "in_key", "%"PRIu64,
491 ntohll(tnl_cfg->in_key));
494 if (tnl_cfg->out_key_flow) {
495 smap_add(args, "out_key", "flow");
496 } else if (tnl_cfg->out_key_present) {
497 smap_add_format(args, "out_key", "%"PRIu64,
498 ntohll(tnl_cfg->out_key));
502 if (tnl_cfg->ttl_inherit) {
503 smap_add(args, "ttl", "inherit");
504 } else if (tnl_cfg->ttl != DEFAULT_TTL) {
505 smap_add_format(args, "ttl", "%"PRIu8, tnl_cfg->ttl);
508 if (tnl_cfg->tos_inherit) {
509 smap_add(args, "tos", "inherit");
510 } else if (tnl_cfg->tos) {
511 smap_add_format(args, "tos", "0x%x", tnl_cfg->tos);
514 if (tnl_cfg->dst_port) {
515 uint16_t dst_port = ntohs(tnl_cfg->dst_port);
516 const char *type = netdev_get_type(dev);
518 if ((!strcmp("vxlan", type) && dst_port != VXLAN_DST_PORT) ||
519 (!strcmp("lisp", type) && dst_port != LISP_DST_PORT)) {
520 smap_add_format(args, "dst_port", "%d", dst_port);
525 smap_add(args, "csum", "true");
528 if (!tnl_cfg->dont_fragment) {
529 smap_add(args, "df_default", "false");
535 /* Code specific to patch ports. */
538 netdev_vport_patch_peer(const struct netdev *netdev)
540 return (netdev_vport_is_patch(netdev)
541 ? netdev_vport_cast(netdev)->peer
546 netdev_vport_inc_rx(const struct netdev *netdev,
547 const struct dpif_flow_stats *stats)
549 if (is_vport_class(netdev_get_class(netdev))) {
550 struct netdev_vport *dev = netdev_vport_cast(netdev);
551 dev->stats.rx_packets += stats->n_packets;
552 dev->stats.rx_bytes += stats->n_bytes;
557 netdev_vport_inc_tx(const struct netdev *netdev,
558 const struct dpif_flow_stats *stats)
560 if (is_vport_class(netdev_get_class(netdev))) {
561 struct netdev_vport *dev = netdev_vport_cast(netdev);
562 dev->stats.tx_packets += stats->n_packets;
563 dev->stats.tx_bytes += stats->n_bytes;
568 get_patch_config(const struct netdev *dev_, struct smap *args)
570 struct netdev_vport *dev = netdev_vport_cast(dev_);
573 smap_add(args, "peer", dev->peer);
579 set_patch_config(struct netdev *dev_, const struct smap *args)
581 struct netdev_vport *dev = netdev_vport_cast(dev_);
582 const char *name = netdev_get_name(dev_);
585 peer = smap_get(args, "peer");
587 VLOG_ERR("%s: patch type requires valid 'peer' argument", name);
591 if (smap_count(args) > 1) {
592 VLOG_ERR("%s: patch type takes only a 'peer' argument", name);
596 if (!strcmp(name, peer)) {
597 VLOG_ERR("%s: patch peer must not be self", name);
602 dev->peer = xstrdup(peer);
608 get_stats(const struct netdev *netdev, struct netdev_stats *stats)
610 struct netdev_vport *dev = netdev_vport_cast(netdev);
611 memcpy(stats, &dev->stats, sizeof *stats);
615 #define VPORT_FUNCTIONS(GET_CONFIG, SET_CONFIG, \
616 GET_TUNNEL_CONFIG, GET_STATUS) \
621 netdev_vport_create, \
622 netdev_vport_destroy, \
627 NULL, /* rx_open */ \
630 NULL, /* send_wait */ \
632 netdev_vport_set_etheraddr, \
633 netdev_vport_get_etheraddr, \
634 NULL, /* get_mtu */ \
635 NULL, /* set_mtu */ \
636 NULL, /* get_ifindex */ \
637 NULL, /* get_carrier */ \
638 NULL, /* get_carrier_resets */ \
639 NULL, /* get_miimon */ \
641 NULL, /* set_stats */ \
643 NULL, /* get_features */ \
644 NULL, /* set_advertisements */ \
646 NULL, /* set_policing */ \
647 NULL, /* get_qos_types */ \
648 NULL, /* get_qos_capabilities */ \
649 NULL, /* get_qos */ \
650 NULL, /* set_qos */ \
651 NULL, /* get_queue */ \
652 NULL, /* set_queue */ \
653 NULL, /* delete_queue */ \
654 NULL, /* get_queue_stats */ \
655 NULL, /* dump_queues */ \
656 NULL, /* dump_queue_stats */ \
658 NULL, /* get_in4 */ \
659 NULL, /* set_in4 */ \
660 NULL, /* get_in6 */ \
661 NULL, /* add_router */ \
662 NULL, /* get_next_hop */ \
664 NULL, /* arp_lookup */ \
666 netdev_vport_update_flags, \
668 netdev_vport_change_seq
670 #define TUNNEL_CLASS(NAME, DPIF_PORT) \
672 { NAME, VPORT_FUNCTIONS(get_tunnel_config, \
674 get_netdev_tunnel_config, \
675 tunnel_get_status) }}
678 netdev_vport_tunnel_register(void)
680 static const struct vport_class vport_classes[] = {
681 TUNNEL_CLASS("gre", "gre_system"),
682 TUNNEL_CLASS("ipsec_gre", "gre_system"),
683 TUNNEL_CLASS("gre64", "gre64_system"),
684 TUNNEL_CLASS("ipsec_gre64", "gre64_system"),
685 TUNNEL_CLASS("vxlan", "vxlan_system"),
686 TUNNEL_CLASS("lisp", "lisp_system")
694 for (i = 0; i < ARRAY_SIZE(vport_classes); i++) {
695 netdev_register_provider(&vport_classes[i].netdev_class);
701 netdev_vport_patch_register(void)
703 static const struct vport_class patch_class =
705 { "patch", VPORT_FUNCTIONS(get_patch_config,
709 netdev_register_provider(&patch_class.netdev_class);