]> git.proxmox.com Git - mirror_ovs.git/blob - lib/netdev-vport.c
vswitch: Remove PMTUD support from userspace.
[mirror_ovs.git] / lib / netdev-vport.c
1 /*
2 * Copyright (c) 2010, 2011, 2012, 2013 Nicira, Inc.
3 *
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:
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
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.
15 */
16
17 #include <config.h>
18
19 #include "netdev-vport.h"
20
21 #include <errno.h>
22 #include <fcntl.h>
23 #include <sys/socket.h>
24 #include <linux/openvswitch.h>
25 #include <linux/rtnetlink.h>
26 #include <net/if.h>
27 #include <sys/ioctl.h>
28
29 #include "byte-order.h"
30 #include "daemon.h"
31 #include "dirs.h"
32 #include "dpif-linux.h"
33 #include "hash.h"
34 #include "hmap.h"
35 #include "list.h"
36 #include "netdev-linux.h"
37 #include "netdev-provider.h"
38 #include "netlink.h"
39 #include "netlink-notifier.h"
40 #include "netlink-socket.h"
41 #include "ofpbuf.h"
42 #include "openvswitch/tunnel.h"
43 #include "packets.h"
44 #include "route-table.h"
45 #include "shash.h"
46 #include "socket-util.h"
47 #include "unaligned.h"
48 #include "vlog.h"
49
50 VLOG_DEFINE_THIS_MODULE(netdev_vport);
51
52 /* Default to the OTV port, per the VXLAN IETF draft. */
53 #define VXLAN_DST_PORT 8472
54
55 struct netdev_dev_vport {
56 struct netdev_dev netdev_dev;
57 struct ofpbuf *options;
58 unsigned int change_seq;
59 };
60
61 struct netdev_vport {
62 struct netdev netdev;
63 };
64
65 struct vport_class {
66 enum ovs_vport_type type;
67 struct netdev_class netdev_class;
68 int (*parse_config)(const char *name, const char *type,
69 const struct smap *args, struct ofpbuf *options);
70 int (*unparse_config)(const char *name, const char *type,
71 const struct nlattr *options, size_t options_len,
72 struct smap *args);
73 };
74
75 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
76
77 static int netdev_vport_create(const struct netdev_class *, const char *,
78 struct netdev_dev **);
79 static void netdev_vport_poll_notify(const struct netdev *);
80 static int tnl_port_config_from_nlattr(const struct nlattr *options,
81 size_t options_len,
82 struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]);
83
84 static bool
85 is_vport_class(const struct netdev_class *class)
86 {
87 return class->create == netdev_vport_create;
88 }
89
90 static const struct vport_class *
91 vport_class_cast(const struct netdev_class *class)
92 {
93 assert(is_vport_class(class));
94 return CONTAINER_OF(class, struct vport_class, netdev_class);
95 }
96
97 static struct netdev_dev_vport *
98 netdev_dev_vport_cast(const struct netdev_dev *netdev_dev)
99 {
100 assert(is_vport_class(netdev_dev_get_class(netdev_dev)));
101 return CONTAINER_OF(netdev_dev, struct netdev_dev_vport, netdev_dev);
102 }
103
104 static struct netdev_dev_vport *
105 netdev_vport_get_dev(const struct netdev *netdev)
106 {
107 return netdev_dev_vport_cast(netdev_get_dev(netdev));
108 }
109
110 static struct netdev_vport *
111 netdev_vport_cast(const struct netdev *netdev)
112 {
113 struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
114 assert(is_vport_class(netdev_dev_get_class(netdev_dev)));
115 return CONTAINER_OF(netdev, struct netdev_vport, netdev);
116 }
117
118 /* If 'netdev' is a vport netdev, returns an ofpbuf that contains Netlink
119 * options to include in OVS_VPORT_ATTR_OPTIONS for configuring that vport.
120 * Otherwise returns NULL. */
121 const struct ofpbuf *
122 netdev_vport_get_options(const struct netdev *netdev)
123 {
124 const struct netdev_dev *dev = netdev_get_dev(netdev);
125
126 return (is_vport_class(netdev_dev_get_class(dev))
127 ? netdev_dev_vport_cast(dev)->options
128 : NULL);
129 }
130
131 enum ovs_vport_type
132 netdev_vport_get_vport_type(const struct netdev *netdev)
133 {
134 const struct netdev_dev *dev = netdev_get_dev(netdev);
135 const struct netdev_class *class = netdev_dev_get_class(dev);
136
137 return (is_vport_class(class) ? vport_class_cast(class)->type
138 : class == &netdev_internal_class ? OVS_VPORT_TYPE_INTERNAL
139 : (class == &netdev_linux_class ||
140 class == &netdev_tap_class) ? OVS_VPORT_TYPE_NETDEV
141 : OVS_VPORT_TYPE_UNSPEC);
142 }
143
144 static uint32_t
145 get_u32_or_zero(const struct nlattr *a)
146 {
147 return a ? nl_attr_get_u32(a) : 0;
148 }
149
150 const char *
151 netdev_vport_get_netdev_type(const struct dpif_linux_vport *vport)
152 {
153 struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1];
154
155 switch (vport->type) {
156 case OVS_VPORT_TYPE_UNSPEC:
157 break;
158
159 case OVS_VPORT_TYPE_NETDEV:
160 return "system";
161
162 case OVS_VPORT_TYPE_INTERNAL:
163 return "internal";
164
165 case OVS_VPORT_TYPE_PATCH:
166 return "patch";
167
168 case OVS_VPORT_TYPE_GRE:
169 if (tnl_port_config_from_nlattr(vport->options, vport->options_len,
170 a)) {
171 break;
172 }
173 return (get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC
174 ? "ipsec_gre" : "gre");
175
176 case OVS_VPORT_TYPE_GRE64:
177 if (tnl_port_config_from_nlattr(vport->options, vport->options_len,
178 a)) {
179 break;
180 }
181 return (get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC
182 ? "ipsec_gre64" : "gre64");
183
184 case OVS_VPORT_TYPE_CAPWAP:
185 return "capwap";
186
187 case OVS_VPORT_TYPE_VXLAN:
188 return "vxlan";
189
190 case OVS_VPORT_TYPE_FT_GRE:
191 case __OVS_VPORT_TYPE_MAX:
192 break;
193 }
194
195 VLOG_WARN_RL(&rl, "dp%d: port `%s' has unsupported type %u",
196 vport->dp_ifindex, vport->name, (unsigned int) vport->type);
197 return "unknown";
198 }
199
200 static int
201 netdev_vport_create(const struct netdev_class *netdev_class, const char *name,
202 struct netdev_dev **netdev_devp)
203 {
204 struct netdev_dev_vport *dev;
205
206 dev = xmalloc(sizeof *dev);
207 netdev_dev_init(&dev->netdev_dev, name, netdev_class);
208 dev->options = NULL;
209 dev->change_seq = 1;
210
211 *netdev_devp = &dev->netdev_dev;
212 route_table_register();
213
214 return 0;
215 }
216
217 static void
218 netdev_vport_destroy(struct netdev_dev *netdev_dev_)
219 {
220 struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_);
221
222 ofpbuf_delete(netdev_dev->options);
223 route_table_unregister();
224 free(netdev_dev);
225 }
226
227 static int
228 netdev_vport_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp)
229 {
230 struct netdev_vport *netdev;
231
232 netdev = xmalloc(sizeof *netdev);
233 netdev_init(&netdev->netdev, netdev_dev_);
234
235 *netdevp = &netdev->netdev;
236 return 0;
237 }
238
239 static void
240 netdev_vport_close(struct netdev *netdev_)
241 {
242 struct netdev_vport *netdev = netdev_vport_cast(netdev_);
243 free(netdev);
244 }
245
246 static int
247 netdev_vport_get_config(struct netdev_dev *dev_, struct smap *args)
248 {
249 const struct netdev_class *netdev_class = netdev_dev_get_class(dev_);
250 const struct vport_class *vport_class = vport_class_cast(netdev_class);
251 struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
252 const char *name = netdev_dev_get_name(dev_);
253 int error;
254
255 if (!dev->options) {
256 struct dpif_linux_vport reply;
257 struct ofpbuf *buf;
258
259 error = dpif_linux_vport_get(name, &reply, &buf);
260 if (error) {
261 VLOG_ERR_RL(&rl, "%s: vport query failed (%s)",
262 name, strerror(error));
263 return error;
264 }
265
266 dev->options = ofpbuf_clone_data(reply.options, reply.options_len);
267 ofpbuf_delete(buf);
268 }
269
270 error = vport_class->unparse_config(name, netdev_class->type,
271 dev->options->data,
272 dev->options->size,
273 args);
274 if (error) {
275 VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)",
276 name, strerror(error));
277 }
278 return error;
279 }
280
281 static int
282 netdev_vport_set_config(struct netdev_dev *dev_, const struct smap *args)
283 {
284 const struct netdev_class *netdev_class = netdev_dev_get_class(dev_);
285 const struct vport_class *vport_class = vport_class_cast(netdev_class);
286 struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
287 const char *name = netdev_dev_get_name(dev_);
288 struct ofpbuf *options;
289 int error;
290
291 options = ofpbuf_new(64);
292 error = vport_class->parse_config(name, netdev_dev_get_type(dev_),
293 args, options);
294 if (!error
295 && (!dev->options
296 || options->size != dev->options->size
297 || memcmp(options->data, dev->options->data, options->size))) {
298 struct dpif_linux_vport vport;
299
300 dpif_linux_vport_init(&vport);
301 vport.cmd = OVS_VPORT_CMD_SET;
302 vport.name = name;
303 vport.options = options->data;
304 vport.options_len = options->size;
305 error = dpif_linux_vport_transact(&vport, NULL, NULL);
306 if (!error || error == ENODEV) {
307 /* Either reconfiguration succeeded or this vport is not installed
308 * in the kernel (e.g. it hasn't been added to a dpif yet with
309 * dpif_port_add()). */
310 ofpbuf_delete(dev->options);
311 dev->options = options;
312 options = NULL;
313 error = 0;
314 }
315 }
316 ofpbuf_delete(options);
317
318 return error;
319 }
320
321 static int
322 netdev_vport_set_etheraddr(struct netdev *netdev,
323 const uint8_t mac[ETH_ADDR_LEN])
324 {
325 struct dpif_linux_vport vport;
326 int error;
327
328 dpif_linux_vport_init(&vport);
329 vport.cmd = OVS_VPORT_CMD_SET;
330 vport.name = netdev_get_name(netdev);
331 vport.address = mac;
332
333 error = dpif_linux_vport_transact(&vport, NULL, NULL);
334 if (!error) {
335 netdev_vport_poll_notify(netdev);
336 }
337 return error;
338 }
339
340 static int
341 netdev_vport_get_etheraddr(const struct netdev *netdev,
342 uint8_t mac[ETH_ADDR_LEN])
343 {
344 struct dpif_linux_vport reply;
345 struct ofpbuf *buf;
346 int error;
347
348 error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf);
349 if (!error) {
350 if (reply.address) {
351 memcpy(mac, reply.address, ETH_ADDR_LEN);
352 } else {
353 error = EOPNOTSUPP;
354 }
355 ofpbuf_delete(buf);
356 }
357 return error;
358 }
359
360 /* Copies 'src' into 'dst', performing format conversion in the process.
361 *
362 * 'src' is allowed to be misaligned. */
363 static void
364 netdev_stats_from_ovs_vport_stats(struct netdev_stats *dst,
365 const struct ovs_vport_stats *src)
366 {
367 dst->rx_packets = get_unaligned_u64(&src->rx_packets);
368 dst->tx_packets = get_unaligned_u64(&src->tx_packets);
369 dst->rx_bytes = get_unaligned_u64(&src->rx_bytes);
370 dst->tx_bytes = get_unaligned_u64(&src->tx_bytes);
371 dst->rx_errors = get_unaligned_u64(&src->rx_errors);
372 dst->tx_errors = get_unaligned_u64(&src->tx_errors);
373 dst->rx_dropped = get_unaligned_u64(&src->rx_dropped);
374 dst->tx_dropped = get_unaligned_u64(&src->tx_dropped);
375 dst->multicast = 0;
376 dst->collisions = 0;
377 dst->rx_length_errors = 0;
378 dst->rx_over_errors = 0;
379 dst->rx_crc_errors = 0;
380 dst->rx_frame_errors = 0;
381 dst->rx_fifo_errors = 0;
382 dst->rx_missed_errors = 0;
383 dst->tx_aborted_errors = 0;
384 dst->tx_carrier_errors = 0;
385 dst->tx_fifo_errors = 0;
386 dst->tx_heartbeat_errors = 0;
387 dst->tx_window_errors = 0;
388 }
389
390 int
391 netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats)
392 {
393 struct dpif_linux_vport reply;
394 struct ofpbuf *buf;
395 int error;
396
397 error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf);
398 if (error) {
399 return error;
400 } else if (!reply.stats) {
401 ofpbuf_delete(buf);
402 return EOPNOTSUPP;
403 }
404
405 netdev_stats_from_ovs_vport_stats(stats, reply.stats);
406
407 ofpbuf_delete(buf);
408
409 return 0;
410 }
411
412 static int
413 tunnel_get_status(const struct netdev *netdev, struct smap *smap)
414 {
415 struct netdev_dev_vport *ndv = netdev_vport_get_dev(netdev);
416 struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1];
417 static char iface[IFNAMSIZ];
418 ovs_be32 route;
419
420 if (!ndv->options) {
421 /* Race condition when 'ndv' was created, but did not have it's
422 * configuration set yet. */
423 return 0;
424 }
425
426 if (tnl_port_config_from_nlattr(ndv->options->data,
427 ndv->options->size, a)) {
428 return 0;
429 }
430 route = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]);
431
432 if (route_table_get_name(route, iface)) {
433 struct netdev *egress_netdev;
434
435 smap_add(smap, "tunnel_egress_iface", iface);
436
437 if (!netdev_open(iface, "system", &egress_netdev)) {
438 smap_add(smap, "tunnel_egress_iface_carrier",
439 netdev_get_carrier(egress_netdev) ? "up" : "down");
440 netdev_close(egress_netdev);
441 }
442 }
443
444 return 0;
445 }
446
447 static int
448 netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED,
449 enum netdev_flags off, enum netdev_flags on OVS_UNUSED,
450 enum netdev_flags *old_flagsp)
451 {
452 if (off & (NETDEV_UP | NETDEV_PROMISC)) {
453 return EOPNOTSUPP;
454 }
455
456 *old_flagsp = NETDEV_UP | NETDEV_PROMISC;
457 return 0;
458 }
459
460 static unsigned int
461 netdev_vport_change_seq(const struct netdev *netdev)
462 {
463 return netdev_vport_get_dev(netdev)->change_seq;
464 }
465
466 static void
467 netdev_vport_run(void)
468 {
469 route_table_run();
470 }
471
472 static void
473 netdev_vport_wait(void)
474 {
475 route_table_wait();
476 }
477 \f
478 /* Helper functions. */
479
480 static void
481 netdev_vport_poll_notify(const struct netdev *netdev)
482 {
483 struct netdev_dev_vport *ndv = netdev_vport_get_dev(netdev);
484
485 ndv->change_seq++;
486 if (!ndv->change_seq) {
487 ndv->change_seq++;
488 }
489 }
490 \f
491 /* Code specific to individual vport types. */
492
493 static void
494 set_key(const struct smap *args, const char *name, uint16_t type,
495 struct ofpbuf *options)
496 {
497 const char *s;
498
499 s = smap_get(args, name);
500 if (!s) {
501 s = smap_get(args, "key");
502 if (!s) {
503 s = "0";
504 }
505 }
506
507 if (!strcmp(s, "flow")) {
508 /* This is the default if no attribute is present. */
509 } else {
510 nl_msg_put_be64(options, type, htonll(strtoull(s, NULL, 0)));
511 }
512 }
513
514 static int
515 parse_tunnel_config(const char *name, const char *type,
516 const struct smap *args, struct ofpbuf *options)
517 {
518 bool is_gre = false;
519 bool is_ipsec = false;
520 bool needs_dst_port = false;
521 bool found_dst_port = false;
522 struct smap_node *node;
523 bool ipsec_mech_set = false;
524 ovs_be32 daddr = htonl(0);
525 ovs_be32 saddr = htonl(0);
526 uint32_t flags;
527
528 if (!strcmp(type, "capwap")) {
529 VLOG_WARN_ONCE("CAPWAP tunnel support is deprecated.");
530 }
531
532 flags = TNL_F_DF_DEFAULT;
533 if (!strcmp(type, "gre") || !strcmp(type, "gre64")) {
534 is_gre = true;
535 } else if (!strcmp(type, "ipsec_gre") || !strcmp(type, "ipsec_gre64")) {
536 is_gre = true;
537 is_ipsec = true;
538 flags |= TNL_F_IPSEC;
539 } else if (!strcmp(type, "vxlan")) {
540 needs_dst_port = true;
541 }
542
543 SMAP_FOR_EACH (node, args) {
544 if (!strcmp(node->key, "remote_ip")) {
545 struct in_addr in_addr;
546 if (lookup_ip(node->value, &in_addr)) {
547 VLOG_WARN("%s: bad %s 'remote_ip'", name, type);
548 } else {
549 daddr = in_addr.s_addr;
550 }
551 } else if (!strcmp(node->key, "local_ip")) {
552 struct in_addr in_addr;
553 if (lookup_ip(node->value, &in_addr)) {
554 VLOG_WARN("%s: bad %s 'local_ip'", name, type);
555 } else {
556 saddr = in_addr.s_addr;
557 }
558 } else if (!strcmp(node->key, "tos")) {
559 if (!strcmp(node->value, "inherit")) {
560 flags |= TNL_F_TOS_INHERIT;
561 } else {
562 char *endptr;
563 int tos;
564 tos = strtol(node->value, &endptr, 0);
565 if (*endptr == '\0' && tos == (tos & IP_DSCP_MASK)) {
566 nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TOS, tos);
567 } else {
568 VLOG_WARN("%s: invalid TOS %s", name, node->value);
569 }
570 }
571 } else if (!strcmp(node->key, "ttl")) {
572 if (!strcmp(node->value, "inherit")) {
573 flags |= TNL_F_TTL_INHERIT;
574 } else {
575 nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TTL, atoi(node->value));
576 }
577 } else if (!strcmp(node->key, "dst_port") && needs_dst_port) {
578 nl_msg_put_u16(options, OVS_TUNNEL_ATTR_DST_PORT,
579 atoi(node->value));
580 found_dst_port = true;
581 } else if (!strcmp(node->key, "csum") && is_gre) {
582 if (!strcmp(node->value, "true")) {
583 flags |= TNL_F_CSUM;
584 }
585 } else if (!strcmp(node->key, "df_inherit")) {
586 if (!strcmp(node->value, "true")) {
587 flags |= TNL_F_DF_INHERIT;
588 }
589 } else if (!strcmp(node->key, "df_default")) {
590 if (!strcmp(node->value, "false")) {
591 flags &= ~TNL_F_DF_DEFAULT;
592 }
593 } else if (!strcmp(node->key, "peer_cert") && is_ipsec) {
594 if (smap_get(args, "certificate")) {
595 ipsec_mech_set = true;
596 } else {
597 const char *use_ssl_cert;
598
599 /* If the "use_ssl_cert" is true, then "certificate" and
600 * "private_key" will be pulled from the SSL table. The
601 * use of this option is strongly discouraged, since it
602 * will like be removed when multiple SSL configurations
603 * are supported by OVS.
604 */
605 use_ssl_cert = smap_get(args, "use_ssl_cert");
606 if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) {
607 VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument",
608 name);
609 return EINVAL;
610 }
611 ipsec_mech_set = true;
612 }
613 } else if (!strcmp(node->key, "psk") && is_ipsec) {
614 ipsec_mech_set = true;
615 } else if (is_ipsec
616 && (!strcmp(node->key, "certificate")
617 || !strcmp(node->key, "private_key")
618 || !strcmp(node->key, "use_ssl_cert"))) {
619 /* Ignore options not used by the netdev. */
620 } else if (!strcmp(node->key, "key") ||
621 !strcmp(node->key, "in_key") ||
622 !strcmp(node->key, "out_key")) {
623 /* Handled separately below. */
624 } else {
625 VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key);
626 }
627 }
628
629 /* Add a default destination port for VXLAN if none specified. */
630 if (needs_dst_port && !found_dst_port) {
631 nl_msg_put_u16(options, OVS_TUNNEL_ATTR_DST_PORT, VXLAN_DST_PORT);
632 }
633
634 if (is_ipsec) {
635 static pid_t pid = 0;
636 if (pid <= 0) {
637 char *file_name = xasprintf("%s/%s", ovs_rundir(),
638 "ovs-monitor-ipsec.pid");
639 pid = read_pidfile(file_name);
640 free(file_name);
641 }
642
643 if (pid < 0) {
644 VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon",
645 name);
646 return EINVAL;
647 }
648
649 if (smap_get(args, "peer_cert") && smap_get(args, "psk")) {
650 VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name);
651 return EINVAL;
652 }
653
654 if (!ipsec_mech_set) {
655 VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument",
656 name);
657 return EINVAL;
658 }
659 }
660
661 set_key(args, "in_key", OVS_TUNNEL_ATTR_IN_KEY, options);
662 set_key(args, "out_key", OVS_TUNNEL_ATTR_OUT_KEY, options);
663
664 if (!daddr) {
665 VLOG_ERR("%s: %s type requires valid 'remote_ip' argument",
666 name, type);
667 return EINVAL;
668 }
669 nl_msg_put_be32(options, OVS_TUNNEL_ATTR_DST_IPV4, daddr);
670
671 if (saddr) {
672 if (ip_is_multicast(daddr)) {
673 VLOG_WARN("%s: remote_ip is multicast, ignoring local_ip", name);
674 } else {
675 nl_msg_put_be32(options, OVS_TUNNEL_ATTR_SRC_IPV4, saddr);
676 }
677 }
678
679 nl_msg_put_u32(options, OVS_TUNNEL_ATTR_FLAGS, flags);
680
681 return 0;
682 }
683
684 static int
685 tnl_port_config_from_nlattr(const struct nlattr *options, size_t options_len,
686 struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1])
687 {
688 static const struct nl_policy ovs_tunnel_policy[] = {
689 [OVS_TUNNEL_ATTR_FLAGS] = { .type = NL_A_U32, .optional = true },
690 [OVS_TUNNEL_ATTR_DST_IPV4] = { .type = NL_A_BE32, .optional = true },
691 [OVS_TUNNEL_ATTR_SRC_IPV4] = { .type = NL_A_BE32, .optional = true },
692 [OVS_TUNNEL_ATTR_IN_KEY] = { .type = NL_A_BE64, .optional = true },
693 [OVS_TUNNEL_ATTR_OUT_KEY] = { .type = NL_A_BE64, .optional = true },
694 [OVS_TUNNEL_ATTR_TOS] = { .type = NL_A_U8, .optional = true },
695 [OVS_TUNNEL_ATTR_TTL] = { .type = NL_A_U8, .optional = true },
696 [OVS_TUNNEL_ATTR_DST_PORT] = { .type = NL_A_U16, .optional = true },
697 };
698 struct ofpbuf buf;
699
700 ofpbuf_use_const(&buf, options, options_len);
701 if (!nl_policy_parse(&buf, 0, ovs_tunnel_policy,
702 a, ARRAY_SIZE(ovs_tunnel_policy))) {
703 return EINVAL;
704 }
705 return 0;
706 }
707
708 static uint64_t
709 get_be64_or_zero(const struct nlattr *a)
710 {
711 return a ? ntohll(nl_attr_get_be64(a)) : 0;
712 }
713
714 static int
715 unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED,
716 const struct nlattr *options, size_t options_len,
717 struct smap *args)
718 {
719 struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1];
720 uint32_t flags;
721 int error;
722
723 error = tnl_port_config_from_nlattr(options, options_len, a);
724 if (error) {
725 return error;
726 }
727
728 if (a[OVS_TUNNEL_ATTR_DST_IPV4]) {
729 ovs_be32 daddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]);
730 smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(daddr));
731 }
732
733 if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) {
734 ovs_be32 saddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]);
735 smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(saddr));
736 }
737
738 if (!a[OVS_TUNNEL_ATTR_IN_KEY] && !a[OVS_TUNNEL_ATTR_OUT_KEY]) {
739 smap_add(args, "key", "flow");
740 } else {
741 uint64_t in_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_IN_KEY]);
742 uint64_t out_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_OUT_KEY]);
743
744 if (in_key && in_key == out_key) {
745 smap_add_format(args, "key", "%"PRIu64, in_key);
746 } else {
747 if (!a[OVS_TUNNEL_ATTR_IN_KEY]) {
748 smap_add(args, "in_key", "flow");
749 } else if (in_key) {
750 smap_add_format(args, "in_key", "%"PRIu64, in_key);
751 }
752
753 if (!a[OVS_TUNNEL_ATTR_OUT_KEY]) {
754 smap_add(args, "out_key", "flow");
755 } else if (out_key) {
756 smap_add_format(args, "out_key", "%"PRIu64, out_key);
757 }
758 }
759 }
760
761 flags = get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]);
762
763 if (flags & TNL_F_TTL_INHERIT) {
764 smap_add(args, "ttl", "inherit");
765 } else if (a[OVS_TUNNEL_ATTR_TTL]) {
766 int ttl = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TTL]);
767 smap_add_format(args, "ttl", "%d", ttl);
768 }
769
770 if (flags & TNL_F_TOS_INHERIT) {
771 smap_add(args, "tos", "inherit");
772 } else if (a[OVS_TUNNEL_ATTR_TOS]) {
773 int tos = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TOS]);
774 smap_add_format(args, "tos", "0x%x", tos);
775 }
776
777 if (a[OVS_TUNNEL_ATTR_DST_PORT]) {
778 uint16_t dst_port = nl_attr_get_u16(a[OVS_TUNNEL_ATTR_DST_PORT]);
779 if (dst_port != VXLAN_DST_PORT) {
780 smap_add_format(args, "dst_port", "%d", dst_port);
781 }
782 }
783
784 if (flags & TNL_F_CSUM) {
785 smap_add(args, "csum", "true");
786 }
787 if (flags & TNL_F_DF_INHERIT) {
788 smap_add(args, "df_inherit", "true");
789 }
790 if (!(flags & TNL_F_DF_DEFAULT)) {
791 smap_add(args, "df_default", "false");
792 }
793
794 return 0;
795 }
796
797 static int
798 parse_patch_config(const char *name, const char *type OVS_UNUSED,
799 const struct smap *args, struct ofpbuf *options)
800 {
801 const char *peer;
802
803 peer = smap_get(args, "peer");
804 if (!peer) {
805 VLOG_ERR("%s: patch type requires valid 'peer' argument", name);
806 return EINVAL;
807 }
808
809 if (smap_count(args) > 1) {
810 VLOG_ERR("%s: patch type takes only a 'peer' argument", name);
811 return EINVAL;
812 }
813
814 if (strlen(peer) >= IFNAMSIZ) {
815 VLOG_ERR("%s: patch 'peer' arg too long", name);
816 return EINVAL;
817 }
818
819 if (!strcmp(name, peer)) {
820 VLOG_ERR("%s: patch peer must not be self", name);
821 return EINVAL;
822 }
823
824 nl_msg_put_string(options, OVS_PATCH_ATTR_PEER, peer);
825
826 return 0;
827 }
828
829 static int
830 unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED,
831 const struct nlattr *options, size_t options_len,
832 struct smap *args)
833 {
834 static const struct nl_policy ovs_patch_policy[] = {
835 [OVS_PATCH_ATTR_PEER] = { .type = NL_A_STRING,
836 .max_len = IFNAMSIZ,
837 .optional = false }
838 };
839
840 struct nlattr *a[ARRAY_SIZE(ovs_patch_policy)];
841 struct ofpbuf buf;
842
843 ofpbuf_use_const(&buf, options, options_len);
844 if (!nl_policy_parse(&buf, 0, ovs_patch_policy,
845 a, ARRAY_SIZE(ovs_patch_policy))) {
846 return EINVAL;
847 }
848
849 smap_add(args, "peer", nl_attr_get_string(a[OVS_PATCH_ATTR_PEER]));
850 return 0;
851 }
852 \f
853 #define VPORT_FUNCTIONS(GET_STATUS) \
854 NULL, \
855 netdev_vport_run, \
856 netdev_vport_wait, \
857 \
858 netdev_vport_create, \
859 netdev_vport_destroy, \
860 netdev_vport_get_config, \
861 netdev_vport_set_config, \
862 \
863 netdev_vport_open, \
864 netdev_vport_close, \
865 \
866 NULL, /* listen */ \
867 NULL, /* recv */ \
868 NULL, /* recv_wait */ \
869 NULL, /* drain */ \
870 \
871 NULL, /* send */ \
872 NULL, /* send_wait */ \
873 \
874 netdev_vport_set_etheraddr, \
875 netdev_vport_get_etheraddr, \
876 NULL, /* get_mtu */ \
877 NULL, /* set_mtu */ \
878 NULL, /* get_ifindex */ \
879 NULL, /* get_carrier */ \
880 NULL, /* get_carrier_resets */ \
881 NULL, /* get_miimon */ \
882 netdev_vport_get_stats, \
883 NULL, /* set_stats */ \
884 \
885 NULL, /* get_features */ \
886 NULL, /* set_advertisements */ \
887 \
888 NULL, /* set_policing */ \
889 NULL, /* get_qos_types */ \
890 NULL, /* get_qos_capabilities */ \
891 NULL, /* get_qos */ \
892 NULL, /* set_qos */ \
893 NULL, /* get_queue */ \
894 NULL, /* set_queue */ \
895 NULL, /* delete_queue */ \
896 NULL, /* get_queue_stats */ \
897 NULL, /* dump_queues */ \
898 NULL, /* dump_queue_stats */ \
899 \
900 NULL, /* get_in4 */ \
901 NULL, /* set_in4 */ \
902 NULL, /* get_in6 */ \
903 NULL, /* add_router */ \
904 NULL, /* get_next_hop */ \
905 GET_STATUS, \
906 NULL, /* arp_lookup */ \
907 \
908 netdev_vport_update_flags, \
909 \
910 netdev_vport_change_seq
911
912 void
913 netdev_vport_register(void)
914 {
915 static const struct vport_class vport_classes[] = {
916 { OVS_VPORT_TYPE_GRE,
917 { "gre", VPORT_FUNCTIONS(tunnel_get_status) },
918 parse_tunnel_config, unparse_tunnel_config },
919
920 { OVS_VPORT_TYPE_GRE,
921 { "ipsec_gre", VPORT_FUNCTIONS(tunnel_get_status) },
922 parse_tunnel_config, unparse_tunnel_config },
923
924 { OVS_VPORT_TYPE_GRE64,
925 { "gre64", VPORT_FUNCTIONS(tunnel_get_status) },
926 parse_tunnel_config, unparse_tunnel_config },
927
928 { OVS_VPORT_TYPE_GRE64,
929 { "ipsec_gre64", VPORT_FUNCTIONS(tunnel_get_status) },
930 parse_tunnel_config, unparse_tunnel_config },
931
932 { OVS_VPORT_TYPE_CAPWAP,
933 { "capwap", VPORT_FUNCTIONS(tunnel_get_status) },
934 parse_tunnel_config, unparse_tunnel_config },
935
936 { OVS_VPORT_TYPE_VXLAN,
937 { "vxlan", VPORT_FUNCTIONS(tunnel_get_status) },
938 parse_tunnel_config, unparse_tunnel_config },
939
940 { OVS_VPORT_TYPE_PATCH,
941 { "patch", VPORT_FUNCTIONS(NULL) },
942 parse_patch_config, unparse_patch_config }
943 };
944
945 int i;
946
947 for (i = 0; i < ARRAY_SIZE(vport_classes); i++) {
948 netdev_register_provider(&vport_classes[i].netdev_class);
949 }
950 }