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
,
125 char namebuf
[], size_t bufsize
)
127 if (netdev_vport_needs_dst_port(netdev
)) {
128 const struct netdev_vport
*vport
= netdev_vport_cast(netdev
);
129 const char *type
= netdev_get_type(netdev
);
132 * Note: IFNAMSIZ is 16 bytes long. The maximum length of a VXLAN
133 * or LISP port name below is 15 or 14 bytes respectively. Still,
134 * assert here on the size of strlen(type) in case that changes
137 BUILD_ASSERT(NETDEV_VPORT_NAME_BUFSIZE
>= IFNAMSIZ
);
138 ovs_assert(strlen(type
) + 10 < IFNAMSIZ
);
139 snprintf(namebuf
, bufsize
, "%s_sys_%d", type
,
140 ntohs(vport
->tnl_cfg
.dst_port
));
143 const struct netdev_class
*class = netdev_get_class(netdev
);
144 const char *dpif_port
= netdev_vport_class_get_dpif_port(class);
145 return dpif_port
? dpif_port
: netdev_get_name(netdev
);
150 netdev_vport_get_dpif_port_strdup(const struct netdev
*netdev
)
152 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
154 return xstrdup(netdev_vport_get_dpif_port(netdev
, namebuf
,
159 netdev_vport_create(const struct netdev_class
*netdev_class
, const char *name
,
160 struct netdev
**netdevp
)
162 struct netdev_vport
*dev
;
164 dev
= xzalloc(sizeof *dev
);
165 netdev_init(&dev
->up
, name
, netdev_class
);
167 eth_addr_random(dev
->etheraddr
);
170 route_table_register();
176 netdev_vport_destroy(struct netdev
*netdev_
)
178 struct netdev_vport
*netdev
= netdev_vport_cast(netdev_
);
180 route_table_unregister();
186 netdev_vport_set_etheraddr(struct netdev
*netdev_
,
187 const uint8_t mac
[ETH_ADDR_LEN
])
189 struct netdev_vport
*netdev
= netdev_vport_cast(netdev_
);
190 memcpy(netdev
->etheraddr
, mac
, ETH_ADDR_LEN
);
191 netdev_vport_poll_notify(netdev
);
196 netdev_vport_get_etheraddr(const struct netdev
*netdev
,
197 uint8_t mac
[ETH_ADDR_LEN
])
199 memcpy(mac
, netdev_vport_cast(netdev
)->etheraddr
, ETH_ADDR_LEN
);
204 tunnel_get_status(const struct netdev
*netdev
, struct smap
*smap
)
206 char iface
[IFNAMSIZ
];
209 route
= netdev_vport_cast(netdev
)->tnl_cfg
.ip_dst
;
210 if (route_table_get_name(route
, iface
)) {
211 struct netdev
*egress_netdev
;
213 smap_add(smap
, "tunnel_egress_iface", iface
);
215 if (!netdev_open(iface
, "system", &egress_netdev
)) {
216 smap_add(smap
, "tunnel_egress_iface_carrier",
217 netdev_get_carrier(egress_netdev
) ? "up" : "down");
218 netdev_close(egress_netdev
);
226 netdev_vport_update_flags(struct netdev
*netdev OVS_UNUSED
,
227 enum netdev_flags off
,
228 enum netdev_flags on OVS_UNUSED
,
229 enum netdev_flags
*old_flagsp
)
231 if (off
& (NETDEV_UP
| NETDEV_PROMISC
)) {
235 *old_flagsp
= NETDEV_UP
| NETDEV_PROMISC
;
240 netdev_vport_change_seq(const struct netdev
*netdev
)
242 return netdev_vport_cast(netdev
)->change_seq
;
246 netdev_vport_run(void)
252 netdev_vport_wait(void)
257 /* Helper functions. */
260 netdev_vport_poll_notify(struct netdev_vport
*ndv
)
263 if (!ndv
->change_seq
) {
268 /* Code specific to tunnel types. */
271 parse_key(const struct smap
*args
, const char *name
,
272 bool *present
, bool *flow
)
279 s
= smap_get(args
, name
);
281 s
= smap_get(args
, "key");
289 if (!strcmp(s
, "flow")) {
293 return htonll(strtoull(s
, NULL
, 0));
298 set_tunnel_config(struct netdev
*dev_
, const struct smap
*args
)
300 struct netdev_vport
*dev
= netdev_vport_cast(dev_
);
301 const char *name
= netdev_get_name(dev_
);
302 const char *type
= netdev_get_type(dev_
);
303 bool ipsec_mech_set
, needs_dst_port
, has_csum
;
304 struct netdev_tunnel_config tnl_cfg
;
305 struct smap_node
*node
;
307 has_csum
= strstr(type
, "gre");
308 ipsec_mech_set
= false;
309 memset(&tnl_cfg
, 0, sizeof tnl_cfg
);
311 needs_dst_port
= netdev_vport_needs_dst_port(dev_
);
312 tnl_cfg
.ipsec
= strstr(type
, "ipsec");
313 tnl_cfg
.dont_fragment
= true;
315 SMAP_FOR_EACH (node
, args
) {
316 if (!strcmp(node
->key
, "remote_ip")) {
317 struct in_addr in_addr
;
318 if (!strcmp(node
->value
, "flow")) {
319 tnl_cfg
.ip_dst_flow
= true;
320 tnl_cfg
.ip_dst
= htonl(0);
321 } else if (lookup_ip(node
->value
, &in_addr
)) {
322 VLOG_WARN("%s: bad %s 'remote_ip'", name
, type
);
323 } else if (ip_is_multicast(in_addr
.s_addr
)) {
324 VLOG_WARN("%s: multicast remote_ip="IP_FMT
" not allowed",
325 name
, IP_ARGS(in_addr
.s_addr
));
328 tnl_cfg
.ip_dst
= in_addr
.s_addr
;
330 } else if (!strcmp(node
->key
, "local_ip")) {
331 struct in_addr in_addr
;
332 if (!strcmp(node
->value
, "flow")) {
333 tnl_cfg
.ip_src_flow
= true;
334 tnl_cfg
.ip_src
= htonl(0);
335 } else if (lookup_ip(node
->value
, &in_addr
)) {
336 VLOG_WARN("%s: bad %s 'local_ip'", name
, type
);
338 tnl_cfg
.ip_src
= in_addr
.s_addr
;
340 } else if (!strcmp(node
->key
, "tos")) {
341 if (!strcmp(node
->value
, "inherit")) {
342 tnl_cfg
.tos_inherit
= true;
346 tos
= strtol(node
->value
, &endptr
, 0);
347 if (*endptr
== '\0' && tos
== (tos
& IP_DSCP_MASK
)) {
350 VLOG_WARN("%s: invalid TOS %s", name
, node
->value
);
353 } else if (!strcmp(node
->key
, "ttl")) {
354 if (!strcmp(node
->value
, "inherit")) {
355 tnl_cfg
.ttl_inherit
= true;
357 tnl_cfg
.ttl
= atoi(node
->value
);
359 } else if (!strcmp(node
->key
, "dst_port") && needs_dst_port
) {
360 tnl_cfg
.dst_port
= htons(atoi(node
->value
));
361 } else if (!strcmp(node
->key
, "csum") && has_csum
) {
362 if (!strcmp(node
->value
, "true")) {
365 } else if (!strcmp(node
->key
, "df_default")) {
366 if (!strcmp(node
->value
, "false")) {
367 tnl_cfg
.dont_fragment
= false;
369 } else if (!strcmp(node
->key
, "peer_cert") && tnl_cfg
.ipsec
) {
370 if (smap_get(args
, "certificate")) {
371 ipsec_mech_set
= true;
373 const char *use_ssl_cert
;
375 /* If the "use_ssl_cert" is true, then "certificate" and
376 * "private_key" will be pulled from the SSL table. The
377 * use of this option is strongly discouraged, since it
378 * will like be removed when multiple SSL configurations
379 * are supported by OVS.
381 use_ssl_cert
= smap_get(args
, "use_ssl_cert");
382 if (!use_ssl_cert
|| strcmp(use_ssl_cert
, "true")) {
383 VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument",
387 ipsec_mech_set
= true;
389 } else if (!strcmp(node
->key
, "psk") && tnl_cfg
.ipsec
) {
390 ipsec_mech_set
= true;
391 } else if (tnl_cfg
.ipsec
392 && (!strcmp(node
->key
, "certificate")
393 || !strcmp(node
->key
, "private_key")
394 || !strcmp(node
->key
, "use_ssl_cert"))) {
395 /* Ignore options not used by the netdev. */
396 } else if (!strcmp(node
->key
, "key") ||
397 !strcmp(node
->key
, "in_key") ||
398 !strcmp(node
->key
, "out_key")) {
399 /* Handled separately below. */
401 VLOG_WARN("%s: unknown %s argument '%s'", name
, type
, node
->key
);
405 /* Add a default destination port for VXLAN if none specified. */
406 if (!strcmp(type
, "vxlan") && !tnl_cfg
.dst_port
) {
407 tnl_cfg
.dst_port
= htons(VXLAN_DST_PORT
);
410 /* Add a default destination port for LISP if none specified. */
411 if (!strcmp(type
, "lisp") && !tnl_cfg
.dst_port
) {
412 tnl_cfg
.dst_port
= htons(LISP_DST_PORT
);
416 static pid_t pid
= 0;
418 char *file_name
= xasprintf("%s/%s", ovs_rundir(),
419 "ovs-monitor-ipsec.pid");
420 pid
= read_pidfile(file_name
);
425 VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon",
430 if (smap_get(args
, "peer_cert") && smap_get(args
, "psk")) {
431 VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name
);
435 if (!ipsec_mech_set
) {
436 VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument",
442 if (!tnl_cfg
.ip_dst
&& !tnl_cfg
.ip_dst_flow
) {
443 VLOG_ERR("%s: %s type requires valid 'remote_ip' argument",
447 if (tnl_cfg
.ip_src_flow
&& !tnl_cfg
.ip_dst_flow
) {
448 VLOG_ERR("%s: %s type requires 'remote_ip=flow' with 'local_ip=flow'",
453 tnl_cfg
.ttl
= DEFAULT_TTL
;
456 tnl_cfg
.in_key
= parse_key(args
, "in_key",
457 &tnl_cfg
.in_key_present
,
458 &tnl_cfg
.in_key_flow
);
460 tnl_cfg
.out_key
= parse_key(args
, "out_key",
461 &tnl_cfg
.out_key_present
,
462 &tnl_cfg
.out_key_flow
);
464 dev
->tnl_cfg
= tnl_cfg
;
465 netdev_vport_poll_notify(dev
);
471 get_tunnel_config(const struct netdev
*dev
, struct smap
*args
)
473 const struct netdev_tunnel_config
*tnl_cfg
=
474 &netdev_vport_cast(dev
)->tnl_cfg
;
476 if (tnl_cfg
->ip_dst
) {
477 smap_add_format(args
, "remote_ip", IP_FMT
, IP_ARGS(tnl_cfg
->ip_dst
));
478 } else if (tnl_cfg
->ip_dst_flow
) {
479 smap_add(args
, "remote_ip", "flow");
482 if (tnl_cfg
->ip_src
) {
483 smap_add_format(args
, "local_ip", IP_FMT
, IP_ARGS(tnl_cfg
->ip_src
));
484 } else if (tnl_cfg
->ip_src_flow
) {
485 smap_add(args
, "local_ip", "flow");
488 if (tnl_cfg
->in_key_flow
&& tnl_cfg
->out_key_flow
) {
489 smap_add(args
, "key", "flow");
490 } else if (tnl_cfg
->in_key_present
&& tnl_cfg
->out_key_present
491 && tnl_cfg
->in_key
== tnl_cfg
->out_key
) {
492 smap_add_format(args
, "key", "%"PRIu64
, ntohll(tnl_cfg
->in_key
));
494 if (tnl_cfg
->in_key_flow
) {
495 smap_add(args
, "in_key", "flow");
496 } else if (tnl_cfg
->in_key_present
) {
497 smap_add_format(args
, "in_key", "%"PRIu64
,
498 ntohll(tnl_cfg
->in_key
));
501 if (tnl_cfg
->out_key_flow
) {
502 smap_add(args
, "out_key", "flow");
503 } else if (tnl_cfg
->out_key_present
) {
504 smap_add_format(args
, "out_key", "%"PRIu64
,
505 ntohll(tnl_cfg
->out_key
));
509 if (tnl_cfg
->ttl_inherit
) {
510 smap_add(args
, "ttl", "inherit");
511 } else if (tnl_cfg
->ttl
!= DEFAULT_TTL
) {
512 smap_add_format(args
, "ttl", "%"PRIu8
, tnl_cfg
->ttl
);
515 if (tnl_cfg
->tos_inherit
) {
516 smap_add(args
, "tos", "inherit");
517 } else if (tnl_cfg
->tos
) {
518 smap_add_format(args
, "tos", "0x%x", tnl_cfg
->tos
);
521 if (tnl_cfg
->dst_port
) {
522 uint16_t dst_port
= ntohs(tnl_cfg
->dst_port
);
523 const char *type
= netdev_get_type(dev
);
525 if ((!strcmp("vxlan", type
) && dst_port
!= VXLAN_DST_PORT
) ||
526 (!strcmp("lisp", type
) && dst_port
!= LISP_DST_PORT
)) {
527 smap_add_format(args
, "dst_port", "%d", dst_port
);
532 smap_add(args
, "csum", "true");
535 if (!tnl_cfg
->dont_fragment
) {
536 smap_add(args
, "df_default", "false");
542 /* Code specific to patch ports. */
545 netdev_vport_patch_peer(const struct netdev
*netdev
)
547 return (netdev_vport_is_patch(netdev
)
548 ? netdev_vport_cast(netdev
)->peer
553 netdev_vport_inc_rx(const struct netdev
*netdev
,
554 const struct dpif_flow_stats
*stats
)
556 if (is_vport_class(netdev_get_class(netdev
))) {
557 struct netdev_vport
*dev
= netdev_vport_cast(netdev
);
558 dev
->stats
.rx_packets
+= stats
->n_packets
;
559 dev
->stats
.rx_bytes
+= stats
->n_bytes
;
564 netdev_vport_inc_tx(const struct netdev
*netdev
,
565 const struct dpif_flow_stats
*stats
)
567 if (is_vport_class(netdev_get_class(netdev
))) {
568 struct netdev_vport
*dev
= netdev_vport_cast(netdev
);
569 dev
->stats
.tx_packets
+= stats
->n_packets
;
570 dev
->stats
.tx_bytes
+= stats
->n_bytes
;
575 get_patch_config(const struct netdev
*dev_
, struct smap
*args
)
577 struct netdev_vport
*dev
= netdev_vport_cast(dev_
);
580 smap_add(args
, "peer", dev
->peer
);
586 set_patch_config(struct netdev
*dev_
, const struct smap
*args
)
588 struct netdev_vport
*dev
= netdev_vport_cast(dev_
);
589 const char *name
= netdev_get_name(dev_
);
592 peer
= smap_get(args
, "peer");
594 VLOG_ERR("%s: patch type requires valid 'peer' argument", name
);
598 if (smap_count(args
) > 1) {
599 VLOG_ERR("%s: patch type takes only a 'peer' argument", name
);
603 if (!strcmp(name
, peer
)) {
604 VLOG_ERR("%s: patch peer must not be self", name
);
609 dev
->peer
= xstrdup(peer
);
610 netdev_vport_poll_notify(dev
);
615 get_stats(const struct netdev
*netdev
, struct netdev_stats
*stats
)
617 struct netdev_vport
*dev
= netdev_vport_cast(netdev
);
618 memcpy(stats
, &dev
->stats
, sizeof *stats
);
622 #define VPORT_FUNCTIONS(GET_CONFIG, SET_CONFIG, \
623 GET_TUNNEL_CONFIG, GET_STATUS) \
628 netdev_vport_create, \
629 netdev_vport_destroy, \
634 NULL, /* rx_open */ \
637 NULL, /* send_wait */ \
639 netdev_vport_set_etheraddr, \
640 netdev_vport_get_etheraddr, \
641 NULL, /* get_mtu */ \
642 NULL, /* set_mtu */ \
643 NULL, /* get_ifindex */ \
644 NULL, /* get_carrier */ \
645 NULL, /* get_carrier_resets */ \
646 NULL, /* get_miimon */ \
648 NULL, /* set_stats */ \
650 NULL, /* get_features */ \
651 NULL, /* set_advertisements */ \
653 NULL, /* set_policing */ \
654 NULL, /* get_qos_types */ \
655 NULL, /* get_qos_capabilities */ \
656 NULL, /* get_qos */ \
657 NULL, /* set_qos */ \
658 NULL, /* get_queue */ \
659 NULL, /* set_queue */ \
660 NULL, /* delete_queue */ \
661 NULL, /* get_queue_stats */ \
662 NULL, /* dump_queues */ \
663 NULL, /* dump_queue_stats */ \
665 NULL, /* get_in4 */ \
666 NULL, /* set_in4 */ \
667 NULL, /* get_in6 */ \
668 NULL, /* add_router */ \
669 NULL, /* get_next_hop */ \
671 NULL, /* arp_lookup */ \
673 netdev_vport_update_flags, \
675 netdev_vport_change_seq
677 #define TUNNEL_CLASS(NAME, DPIF_PORT) \
679 { NAME, VPORT_FUNCTIONS(get_tunnel_config, \
681 get_netdev_tunnel_config, \
682 tunnel_get_status) }}
685 netdev_vport_tunnel_register(void)
687 static const struct vport_class vport_classes
[] = {
688 TUNNEL_CLASS("gre", "gre_system"),
689 TUNNEL_CLASS("ipsec_gre", "gre_system"),
690 TUNNEL_CLASS("gre64", "gre64_system"),
691 TUNNEL_CLASS("ipsec_gre64", "gre64_system"),
692 TUNNEL_CLASS("vxlan", "vxlan_system"),
693 TUNNEL_CLASS("lisp", "lisp_system")
701 for (i
= 0; i
< ARRAY_SIZE(vport_classes
); i
++) {
702 netdev_register_provider(&vport_classes
[i
].netdev_class
);
708 netdev_vport_patch_register(void)
710 static const struct vport_class patch_class
=
712 { "patch", VPORT_FUNCTIONS(get_patch_config
,
716 netdev_register_provider(&patch_class
.netdev_class
);