]> git.proxmox.com Git - mirror_ovs.git/blame - lib/netdev-native-tnl.c
ovs-vswitchd: Better document that ovs-vswitchd manages its own datapaths.
[mirror_ovs.git] / lib / netdev-native-tnl.c
CommitLineData
6b241d64
PS
1/*
2 * Copyright (c) 2016 Nicira, Inc.
68da36fe 3 * Copyright (c) 2016 Red Hat, Inc.
6b241d64
PS
4 *
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at:
8 *
9 * http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 */
17
18#include <config.h>
19
aca40d4f
TLSC
20#include "netdev-native-tnl.h"
21
6b241d64
PS
22#include <errno.h>
23#include <fcntl.h>
24#include <sys/socket.h>
25#include <net/if.h>
b2befd5b 26#include <sys/types.h>
67eaddc0 27#include <netinet/in.h>
aca40d4f 28#include <netinet/ip.h>
6b241d64
PS
29#include <netinet/ip6.h>
30#include <sys/ioctl.h>
31
32#include <errno.h>
33#include <stdlib.h>
34#include <sys/time.h>
35
6b241d64
PS
36#include "byte-order.h"
37#include "csum.h"
6b241d64 38#include "dp-packet.h"
aca40d4f 39#include "netdev.h"
6b241d64
PS
40#include "netdev-vport.h"
41#include "netdev-vport-private.h"
42#include "odp-netlink.h"
6b241d64 43#include "packets.h"
aca40d4f 44#include "seq.h"
6b241d64
PS
45#include "unaligned.h"
46#include "unixctl.h"
aca40d4f 47#include "openvswitch/vlog.h"
6b241d64
PS
48
49VLOG_DEFINE_THIS_MODULE(native_tnl);
50static struct vlog_rate_limit err_rl = VLOG_RATE_LIMIT_INIT(60, 5);
51
52#define VXLAN_HLEN (sizeof(struct udp_header) + \
53 sizeof(struct vxlanhdr))
54
55#define GENEVE_BASE_HLEN (sizeof(struct udp_header) + \
56 sizeof(struct genevehdr))
57
58uint16_t tnl_udp_port_min = 32768;
59uint16_t tnl_udp_port_max = 61000;
60
61void *
62netdev_tnl_ip_extract_tnl_md(struct dp_packet *packet, struct flow_tnl *tnl,
63 unsigned int *hlen)
64{
65 void *nh;
66 struct ip_header *ip;
67 struct ovs_16aligned_ip6_hdr *ip6;
68 void *l4;
69 int l3_size;
70
71 nh = dp_packet_l3(packet);
72 ip = nh;
73 ip6 = nh;
74 l4 = dp_packet_l4(packet);
75
76 if (!nh || !l4) {
77 return NULL;
78 }
79
80 *hlen = sizeof(struct eth_header);
81
82 l3_size = dp_packet_size(packet) -
83 ((char *)nh - (char *)dp_packet_data(packet));
84
85 if (IP_VER(ip->ip_ihl_ver) == 4) {
86
87 ovs_be32 ip_src, ip_dst;
88
1a2bb118
SC
89 if (OVS_UNLIKELY(!dp_packet_ip_checksum_valid(packet))) {
90 if (csum(ip, IP_IHL(ip->ip_ihl_ver) * 4)) {
91 VLOG_WARN_RL(&err_rl, "ip packet has invalid checksum");
92 return NULL;
93 }
6b241d64
PS
94 }
95
96 if (ntohs(ip->ip_tot_len) > l3_size) {
97 VLOG_WARN_RL(&err_rl, "ip packet is truncated (IP length %d, actual %d)",
98 ntohs(ip->ip_tot_len), l3_size);
99 return NULL;
100 }
101 if (IP_IHL(ip->ip_ihl_ver) * 4 > sizeof(struct ip_header)) {
102 VLOG_WARN_RL(&err_rl, "ip options not supported on tunnel packets "
103 "(%d bytes)", IP_IHL(ip->ip_ihl_ver) * 4);
104 return NULL;
105 }
106
107 ip_src = get_16aligned_be32(&ip->ip_src);
108 ip_dst = get_16aligned_be32(&ip->ip_dst);
109
110 tnl->ip_src = ip_src;
111 tnl->ip_dst = ip_dst;
112 tnl->ip_tos = ip->ip_tos;
113 tnl->ip_ttl = ip->ip_ttl;
114
115 *hlen += IP_HEADER_LEN;
116
117 } else if (IP_VER(ip->ip_ihl_ver) == 6) {
98c086db 118 ovs_be32 tc_flow = get_16aligned_be32(&ip6->ip6_flow);
6b241d64
PS
119
120 memcpy(tnl->ipv6_src.s6_addr, ip6->ip6_src.be16, sizeof ip6->ip6_src);
121 memcpy(tnl->ipv6_dst.s6_addr, ip6->ip6_dst.be16, sizeof ip6->ip6_dst);
98c086db
PS
122
123 tnl->ip_tos = ntohl(tc_flow) >> 20;
6b241d64
PS
124 tnl->ip_ttl = ip6->ip6_hlim;
125
3456684e 126 *hlen += packet->l4_ofs - packet->l3_ofs;
6b241d64
PS
127
128 } else {
129 VLOG_WARN_RL(&err_rl, "ipv4 packet has invalid version (%d)",
130 IP_VER(ip->ip_ihl_ver));
131 return NULL;
132 }
133
134 return l4;
135}
136
137/* Pushes the 'size' bytes of 'header' into the headroom of 'packet',
138 * reallocating the packet if necessary. 'header' should contain an Ethernet
139 * header, followed by an IPv4 header (without options), and an L4 header.
140 *
141 * This function sets the IP header's ip_tot_len field (which should be zeroed
142 * as part of 'header') and puts its value into '*ip_tot_size' as well. Also
7c12dfc5 143 * updates IP header checksum, as well as the l3 and l4 offsets in 'packet'.
6b241d64
PS
144 *
145 * Return pointer to the L4 header added to 'packet'. */
146void *
147netdev_tnl_push_ip_header(struct dp_packet *packet,
148 const void *header, int size, int *ip_tot_size)
149{
150 struct eth_header *eth;
151 struct ip_header *ip;
152 struct ovs_16aligned_ip6_hdr *ip6;
153
154 eth = dp_packet_push_uninit(packet, size);
155 *ip_tot_size = dp_packet_size(packet) - sizeof (struct eth_header);
156
157 memcpy(eth, header, size);
63171f04
JS
158 /* The encapsulated packet has type Ethernet. Adjust dp_packet. */
159 packet->packet_type = htonl(PT_ETH);
160 dp_packet_reset_offsets(packet);
161 packet->l3_ofs = sizeof (struct eth_header);
6b241d64
PS
162
163 if (netdev_tnl_is_header_ipv6(header)) {
164 ip6 = netdev_tnl_ipv6_hdr(eth);
165 *ip_tot_size -= IPV6_HEADER_LEN;
166 ip6->ip6_plen = htons(*ip_tot_size);
ce8bbd37 167 packet->l4_ofs = dp_packet_size(packet) - *ip_tot_size;
6b241d64
PS
168 return ip6 + 1;
169 } else {
170 ip = netdev_tnl_ip_hdr(eth);
171 ip->ip_tot_len = htons(*ip_tot_size);
172 ip->ip_csum = recalc_csum16(ip->ip_csum, 0, ip->ip_tot_len);
173 *ip_tot_size -= IP_HEADER_LEN;
ce8bbd37 174 packet->l4_ofs = dp_packet_size(packet) - *ip_tot_size;
6b241d64
PS
175 return ip + 1;
176 }
177}
178
179static void *
180udp_extract_tnl_md(struct dp_packet *packet, struct flow_tnl *tnl,
181 unsigned int *hlen)
182{
183 struct udp_header *udp;
184
185 udp = netdev_tnl_ip_extract_tnl_md(packet, tnl, hlen);
186 if (!udp) {
187 return NULL;
188 }
189
190 if (udp->udp_csum) {
1a2bb118
SC
191 if (OVS_UNLIKELY(!dp_packet_l4_checksum_valid(packet))) {
192 uint32_t csum;
193 if (netdev_tnl_is_header_ipv6(dp_packet_data(packet))) {
194 csum = packet_csum_pseudoheader6(dp_packet_l3(packet));
195 } else {
196 csum = packet_csum_pseudoheader(dp_packet_l3(packet));
197 }
198
199 csum = csum_continue(csum, udp, dp_packet_size(packet) -
200 ((const unsigned char *)udp -
2482b0b0 201 (const unsigned char *)dp_packet_eth(packet)
1a2bb118
SC
202 ));
203 if (csum_finish(csum)) {
204 return NULL;
205 }
6b241d64
PS
206 }
207 tnl->flags |= FLOW_TNL_F_CSUM;
208 }
209
210 tnl->tp_src = udp->udp_src;
211 tnl->tp_dst = udp->udp_dst;
212
213 return udp + 1;
214}
215
216
217void
218netdev_tnl_push_udp_header(struct dp_packet *packet,
219 const struct ovs_action_push_tnl *data)
220{
221 struct udp_header *udp;
222 int ip_tot_size;
223
224 udp = netdev_tnl_push_ip_header(packet, data->header, data->header_len, &ip_tot_size);
225
226 /* set udp src port */
227 udp->udp_src = netdev_tnl_get_src_port(packet);
228 udp->udp_len = htons(ip_tot_size);
229
230 if (udp->udp_csum) {
231 uint32_t csum;
232 if (netdev_tnl_is_header_ipv6(dp_packet_data(packet))) {
233 csum = packet_csum_pseudoheader6(netdev_tnl_ipv6_hdr(dp_packet_data(packet)));
234 } else {
235 csum = packet_csum_pseudoheader(netdev_tnl_ip_hdr(dp_packet_data(packet)));
236 }
237
238 csum = csum_continue(csum, udp, ip_tot_size);
239 udp->udp_csum = csum_finish(csum);
240
241 if (!udp->udp_csum) {
242 udp->udp_csum = htons(0xffff);
243 }
244 }
245}
246
247static void *
4975aa3e
PS
248eth_build_header(struct ovs_action_push_tnl *data,
249 const struct netdev_tnl_build_header_params *params)
6b241d64 250{
4975aa3e
PS
251 uint16_t eth_proto = params->is_ipv6 ? ETH_TYPE_IPV6 : ETH_TYPE_IP;
252 struct eth_header *eth;
6b241d64 253
4975aa3e 254 memset(data->header, 0, sizeof data->header);
6b241d64 255
4975aa3e
PS
256 eth = (struct eth_header *)data->header;
257 eth->eth_dst = params->dmac;
258 eth->eth_src = params->smac;
259 eth->eth_type = htons(eth_proto);
260 data->header_len = sizeof(struct eth_header);
261 return eth + 1;
262}
6b241d64 263
4975aa3e
PS
264void *
265netdev_tnl_ip_build_header(struct ovs_action_push_tnl *data,
266 const struct netdev_tnl_build_header_params *params,
267 uint8_t next_proto)
268{
269 void *l3;
270
271 l3 = eth_build_header(data, params);
272 if (!params->is_ipv6) {
273 ovs_be32 ip_src = in6_addr_get_mapped_ipv4(params->s_ip);
274 struct ip_header *ip;
275
276 ip = (struct ip_header *) l3;
277
278 ip->ip_ihl_ver = IP_IHL_VER(5, 4);
279 ip->ip_tos = params->flow->tunnel.ip_tos;
280 ip->ip_ttl = params->flow->tunnel.ip_ttl;
281 ip->ip_proto = next_proto;
282 put_16aligned_be32(&ip->ip_src, ip_src);
283 put_16aligned_be32(&ip->ip_dst, params->flow->tunnel.ip_dst);
284
285 ip->ip_frag_off = (params->flow->tunnel.flags & FLOW_TNL_F_DONT_FRAGMENT) ?
286 htons(IP_DF) : 0;
287
ece9c294 288 /* Checksum has already been zeroed by eth_build_header. */
4975aa3e
PS
289 ip->ip_csum = csum(ip, sizeof *ip);
290
291 data->header_len += IP_HEADER_LEN;
292 return ip + 1;
6b241d64 293 } else {
4975aa3e
PS
294 struct ovs_16aligned_ip6_hdr *ip6;
295
296 ip6 = (struct ovs_16aligned_ip6_hdr *) l3;
297
98c086db
PS
298 put_16aligned_be32(&ip6->ip6_flow, htonl(6 << 28) |
299 htonl(params->flow->tunnel.ip_tos << 20));
4975aa3e
PS
300 ip6->ip6_hlim = params->flow->tunnel.ip_ttl;
301 ip6->ip6_nxt = next_proto;
302 memcpy(&ip6->ip6_src, params->s_ip, sizeof(ovs_be32[4]));
303 memcpy(&ip6->ip6_dst, &params->flow->tunnel.ipv6_dst, sizeof(ovs_be32[4]));
304
305 data->header_len += IPV6_HEADER_LEN;
306 return ip6 + 1;
6b241d64 307 }
4975aa3e
PS
308}
309
310static void *
311udp_build_header(struct netdev_tunnel_config *tnl_cfg,
312 struct ovs_action_push_tnl *data,
313 const struct netdev_tnl_build_header_params *params)
314{
315 struct udp_header *udp;
6b241d64 316
4975aa3e 317 udp = netdev_tnl_ip_build_header(data, params, IPPROTO_UDP);
6b241d64
PS
318 udp->udp_dst = tnl_cfg->dst_port;
319
4975aa3e 320 if (params->is_ipv6 || params->flow->tunnel.flags & FLOW_TNL_F_CSUM) {
6b241d64
PS
321 /* Write a value in now to mark that we should compute the checksum
322 * later. 0xffff is handy because it is transparent to the
323 * calculation. */
324 udp->udp_csum = htons(0xffff);
325 }
4975aa3e 326 data->header_len += sizeof *udp;
6b241d64
PS
327 return udp + 1;
328}
329
330static int
331gre_header_len(ovs_be16 flags)
332{
333 int hlen = 4;
334
335 if (flags & htons(GRE_CSUM)) {
336 hlen += 4;
337 }
338 if (flags & htons(GRE_KEY)) {
339 hlen += 4;
340 }
341 if (flags & htons(GRE_SEQ)) {
342 hlen += 4;
343 }
344 return hlen;
345}
346
347static int
348parse_gre_header(struct dp_packet *packet,
349 struct flow_tnl *tnl)
350{
351 const struct gre_base_hdr *greh;
352 ovs_16aligned_be32 *options;
353 int hlen;
354 unsigned int ulen;
63171f04 355 uint16_t greh_protocol;
6b241d64
PS
356
357 greh = netdev_tnl_ip_extract_tnl_md(packet, tnl, &ulen);
358 if (!greh) {
359 return -EINVAL;
360 }
361
362 if (greh->flags & ~(htons(GRE_CSUM | GRE_KEY | GRE_SEQ))) {
363 return -EINVAL;
364 }
365
6b241d64
PS
366 hlen = ulen + gre_header_len(greh->flags);
367 if (hlen > dp_packet_size(packet)) {
368 return -EINVAL;
369 }
370
371 options = (ovs_16aligned_be32 *)(greh + 1);
372 if (greh->flags & htons(GRE_CSUM)) {
373 ovs_be16 pkt_csum;
374
375 pkt_csum = csum(greh, dp_packet_size(packet) -
376 ((const unsigned char *)greh -
2482b0b0 377 (const unsigned char *)dp_packet_eth(packet)));
6b241d64
PS
378 if (pkt_csum) {
379 return -EINVAL;
380 }
381 tnl->flags = FLOW_TNL_F_CSUM;
382 options++;
383 }
384
385 if (greh->flags & htons(GRE_KEY)) {
3d75c660 386 tnl->tun_id = be32_to_be64(get_16aligned_be32(options));
6b241d64
PS
387 tnl->flags |= FLOW_TNL_F_KEY;
388 options++;
389 }
390
391 if (greh->flags & htons(GRE_SEQ)) {
392 options++;
393 }
394
63171f04
JS
395 /* Set the new packet type depending on the GRE protocol field. */
396 greh_protocol = ntohs(greh->protocol);
397 if (greh_protocol == ETH_TYPE_TEB) {
398 packet->packet_type = htonl(PT_ETH);
399 } else if (greh_protocol >= ETH_TYPE_MIN) {
400 /* Allow all GRE protocol values above 0x5ff as Ethertypes. */
401 packet->packet_type = PACKET_TYPE_BE(OFPHTN_ETHERTYPE, greh_protocol);
402 } else {
403 return -EINVAL;
404 }
405
6b241d64
PS
406 return hlen;
407}
408
1c8f98d9 409struct dp_packet *
6b241d64
PS
410netdev_gre_pop_header(struct dp_packet *packet)
411{
412 struct pkt_metadata *md = &packet->md;
413 struct flow_tnl *tnl = &md->tunnel;
414 int hlen = sizeof(struct eth_header) + 4;
415
416 hlen += netdev_tnl_is_header_ipv6(dp_packet_data(packet)) ?
417 IPV6_HEADER_LEN : IP_HEADER_LEN;
418
419 pkt_metadata_init_tnl(md);
420 if (hlen > dp_packet_size(packet)) {
1c8f98d9 421 goto err;
6b241d64
PS
422 }
423
424 hlen = parse_gre_header(packet, tnl);
425 if (hlen < 0) {
1c8f98d9 426 goto err;
6b241d64
PS
427 }
428
429 dp_packet_reset_packet(packet, hlen);
430
1c8f98d9
PS
431 return packet;
432err:
433 dp_packet_delete(packet);
434 return NULL;
6b241d64
PS
435}
436
437void
438netdev_gre_push_header(struct dp_packet *packet,
439 const struct ovs_action_push_tnl *data)
440{
441 struct gre_base_hdr *greh;
442 int ip_tot_size;
443
444 greh = netdev_tnl_push_ip_header(packet, data->header, data->header_len, &ip_tot_size);
445
446 if (greh->flags & htons(GRE_CSUM)) {
447 ovs_be16 *csum_opt = (ovs_be16 *) (greh + 1);
448 *csum_opt = csum(greh, ip_tot_size);
449 }
450}
451
452int
453netdev_gre_build_header(const struct netdev *netdev,
454 struct ovs_action_push_tnl *data,
4975aa3e 455 const struct netdev_tnl_build_header_params *params)
6b241d64
PS
456{
457 struct netdev_vport *dev = netdev_vport_cast(netdev);
458 struct netdev_tunnel_config *tnl_cfg;
6b241d64
PS
459 struct gre_base_hdr *greh;
460 ovs_16aligned_be32 *options;
4975aa3e 461 unsigned int hlen;
6b241d64
PS
462
463 /* XXX: RCUfy tnl_cfg. */
464 ovs_mutex_lock(&dev->mutex);
465 tnl_cfg = &dev->tnl_cfg;
466
4975aa3e 467 greh = netdev_tnl_ip_build_header(data, params, IPPROTO_GRE);
6b241d64 468
875ab130 469 if (params->flow->packet_type == htonl(PT_ETH)) {
63171f04 470 greh->protocol = htons(ETH_TYPE_TEB);
875ab130
BP
471 } else if (pt_ns(params->flow->packet_type) == OFPHTN_ETHERTYPE) {
472 greh->protocol = pt_ns_type_be(params->flow->packet_type);
473 } else {
474 ovs_mutex_unlock(&dev->mutex);
475 return 1;
63171f04 476 }
6b241d64
PS
477 greh->flags = 0;
478
479 options = (ovs_16aligned_be32 *) (greh + 1);
4975aa3e 480 if (params->flow->tunnel.flags & FLOW_TNL_F_CSUM) {
6b241d64
PS
481 greh->flags |= htons(GRE_CSUM);
482 put_16aligned_be32(options, 0);
483 options++;
484 }
485
486 if (tnl_cfg->out_key_present) {
487 greh->flags |= htons(GRE_KEY);
3d75c660 488 put_16aligned_be32(options, be64_to_be32(params->flow->tunnel.tun_id));
6b241d64
PS
489 options++;
490 }
491
492 ovs_mutex_unlock(&dev->mutex);
493
494 hlen = (uint8_t *) options - (uint8_t *) greh;
495
4975aa3e 496 data->header_len += hlen;
6b241d64
PS
497 data->tnl_type = OVS_VPORT_TYPE_GRE;
498 return 0;
499}
500
1c8f98d9 501struct dp_packet *
6b241d64
PS
502netdev_vxlan_pop_header(struct dp_packet *packet)
503{
504 struct pkt_metadata *md = &packet->md;
505 struct flow_tnl *tnl = &md->tunnel;
506 struct vxlanhdr *vxh;
507 unsigned int hlen;
439f39cb
GS
508 ovs_be32 vx_flags;
509 enum packet_type next_pt = PT_ETH;
6b241d64 510
1fc2e1bd
BB
511 ovs_assert(packet->l3_ofs > 0);
512 ovs_assert(packet->l4_ofs > 0);
513
6b241d64
PS
514 pkt_metadata_init_tnl(md);
515 if (VXLAN_HLEN > dp_packet_l4_size(packet)) {
1c8f98d9 516 goto err;
6b241d64
PS
517 }
518
519 vxh = udp_extract_tnl_md(packet, tnl, &hlen);
520 if (!vxh) {
1c8f98d9 521 goto err;
6b241d64
PS
522 }
523
439f39cb
GS
524 vx_flags = get_16aligned_be32(&vxh->vx_flags);
525 if (vx_flags & htonl(VXLAN_HF_GPE)) {
526 vx_flags &= htonl(~VXLAN_GPE_USED_BITS);
527 /* Drop the OAM packets */
528 if (vxh->vx_gpe.flags & VXLAN_GPE_FLAGS_O) {
529 goto err;
530 }
531 switch (vxh->vx_gpe.next_protocol) {
532 case VXLAN_GPE_NP_IPV4:
533 next_pt = PT_IPV4;
534 break;
535 case VXLAN_GPE_NP_IPV6:
536 next_pt = PT_IPV6;
537 break;
478b1473
JS
538 case VXLAN_GPE_NP_NSH:
539 next_pt = PT_NSH;
540 break;
439f39cb
GS
541 case VXLAN_GPE_NP_ETHERNET:
542 next_pt = PT_ETH;
543 break;
544 default:
545 goto err;
546 }
547 }
548
549 if (vx_flags != htonl(VXLAN_FLAGS) ||
6b241d64
PS
550 (get_16aligned_be32(&vxh->vx_vni) & htonl(0xff))) {
551 VLOG_WARN_RL(&err_rl, "invalid vxlan flags=%#x vni=%#x\n",
439f39cb 552 ntohl(vx_flags),
6b241d64 553 ntohl(get_16aligned_be32(&vxh->vx_vni)));
1c8f98d9 554 goto err;
6b241d64
PS
555 }
556 tnl->tun_id = htonll(ntohl(get_16aligned_be32(&vxh->vx_vni)) >> 8);
557 tnl->flags |= FLOW_TNL_F_KEY;
558
439f39cb 559 packet->packet_type = htonl(next_pt);
6b241d64 560 dp_packet_reset_packet(packet, hlen + VXLAN_HLEN);
439f39cb
GS
561 if (next_pt != PT_ETH) {
562 packet->l3_ofs = 0;
563 }
6b241d64 564
1c8f98d9
PS
565 return packet;
566err:
567 dp_packet_delete(packet);
568 return NULL;
6b241d64
PS
569}
570
571int
572netdev_vxlan_build_header(const struct netdev *netdev,
573 struct ovs_action_push_tnl *data,
4975aa3e 574 const struct netdev_tnl_build_header_params *params)
6b241d64
PS
575{
576 struct netdev_vport *dev = netdev_vport_cast(netdev);
577 struct netdev_tunnel_config *tnl_cfg;
578 struct vxlanhdr *vxh;
6b241d64
PS
579
580 /* XXX: RCUfy tnl_cfg. */
581 ovs_mutex_lock(&dev->mutex);
582 tnl_cfg = &dev->tnl_cfg;
583
4975aa3e 584 vxh = udp_build_header(tnl_cfg, data, params);
6b241d64 585
439f39cb
GS
586 if (tnl_cfg->exts & (1 << OVS_VXLAN_EXT_GPE)) {
587 put_16aligned_be32(&vxh->vx_flags, htonl(VXLAN_FLAGS | VXLAN_HF_GPE));
588 put_16aligned_be32(&vxh->vx_vni,
589 htonl(ntohll(params->flow->tunnel.tun_id) << 8));
875ab130
BP
590 if (params->flow->packet_type == htonl(PT_ETH)) {
591 vxh->vx_gpe.next_protocol = VXLAN_GPE_NP_ETHERNET;
592 } else if (pt_ns(params->flow->packet_type) == OFPHTN_ETHERTYPE) {
593 switch (pt_ns_type(params->flow->packet_type)) {
439f39cb
GS
594 case ETH_TYPE_IP:
595 vxh->vx_gpe.next_protocol = VXLAN_GPE_NP_IPV4;
596 break;
597 case ETH_TYPE_IPV6:
598 vxh->vx_gpe.next_protocol = VXLAN_GPE_NP_IPV6;
599 break;
478b1473
JS
600 case ETH_TYPE_NSH:
601 vxh->vx_gpe.next_protocol = VXLAN_GPE_NP_NSH;
602 break;
439f39cb
GS
603 case ETH_TYPE_TEB:
604 vxh->vx_gpe.next_protocol = VXLAN_GPE_NP_ETHERNET;
605 break;
875ab130
BP
606 default:
607 goto drop;
439f39cb
GS
608 }
609 } else {
875ab130 610 goto drop;
439f39cb
GS
611 }
612 } else {
613 put_16aligned_be32(&vxh->vx_flags, htonl(VXLAN_FLAGS));
614 put_16aligned_be32(&vxh->vx_vni,
615 htonl(ntohll(params->flow->tunnel.tun_id) << 8));
616 }
6b241d64
PS
617
618 ovs_mutex_unlock(&dev->mutex);
4975aa3e 619 data->header_len += sizeof *vxh;
6b241d64
PS
620 data->tnl_type = OVS_VPORT_TYPE_VXLAN;
621 return 0;
875ab130
BP
622
623drop:
624 ovs_mutex_unlock(&dev->mutex);
625 return 1;
6b241d64
PS
626}
627
1c8f98d9 628struct dp_packet *
6b241d64
PS
629netdev_geneve_pop_header(struct dp_packet *packet)
630{
631 struct pkt_metadata *md = &packet->md;
632 struct flow_tnl *tnl = &md->tunnel;
633 struct genevehdr *gnh;
634 unsigned int hlen, opts_len, ulen;
635
636 pkt_metadata_init_tnl(md);
637 if (GENEVE_BASE_HLEN > dp_packet_l4_size(packet)) {
638 VLOG_WARN_RL(&err_rl, "geneve packet too small: min header=%u packet size=%"PRIuSIZE"\n",
639 (unsigned int)GENEVE_BASE_HLEN, dp_packet_l4_size(packet));
1c8f98d9 640 goto err;
6b241d64
PS
641 }
642
643 gnh = udp_extract_tnl_md(packet, tnl, &ulen);
644 if (!gnh) {
1c8f98d9 645 goto err;
6b241d64
PS
646 }
647
648 opts_len = gnh->opt_len * 4;
649 hlen = ulen + GENEVE_BASE_HLEN + opts_len;
650 if (hlen > dp_packet_size(packet)) {
651 VLOG_WARN_RL(&err_rl, "geneve packet too small: header len=%u packet size=%u\n",
652 hlen, dp_packet_size(packet));
1c8f98d9 653 goto err;
6b241d64
PS
654 }
655
656 if (gnh->ver != 0) {
657 VLOG_WARN_RL(&err_rl, "unknown geneve version: %"PRIu8"\n", gnh->ver);
1c8f98d9 658 goto err;
6b241d64
PS
659 }
660
661 if (gnh->proto_type != htons(ETH_TYPE_TEB)) {
662 VLOG_WARN_RL(&err_rl, "unknown geneve encapsulated protocol: %#x\n",
663 ntohs(gnh->proto_type));
1c8f98d9 664 goto err;
6b241d64
PS
665 }
666
667 tnl->flags |= gnh->oam ? FLOW_TNL_F_OAM : 0;
668 tnl->tun_id = htonll(ntohl(get_16aligned_be32(&gnh->vni)) >> 8);
669 tnl->flags |= FLOW_TNL_F_KEY;
670
671 memcpy(tnl->metadata.opts.gnv, gnh->options, opts_len);
672 tnl->metadata.present.len = opts_len;
673 tnl->flags |= FLOW_TNL_F_UDPIF;
674
63171f04 675 packet->packet_type = htonl(PT_ETH);
6b241d64
PS
676 dp_packet_reset_packet(packet, hlen);
677
1c8f98d9
PS
678 return packet;
679err:
680 dp_packet_delete(packet);
681 return NULL;
6b241d64
PS
682}
683
684int
685netdev_geneve_build_header(const struct netdev *netdev,
686 struct ovs_action_push_tnl *data,
4975aa3e 687 const struct netdev_tnl_build_header_params *params)
6b241d64
PS
688{
689 struct netdev_vport *dev = netdev_vport_cast(netdev);
690 struct netdev_tunnel_config *tnl_cfg;
691 struct genevehdr *gnh;
692 int opt_len;
693 bool crit_opt;
6b241d64
PS
694
695 /* XXX: RCUfy tnl_cfg. */
696 ovs_mutex_lock(&dev->mutex);
697 tnl_cfg = &dev->tnl_cfg;
698
4975aa3e 699 gnh = udp_build_header(tnl_cfg, data, params);
6b241d64 700
4975aa3e 701 put_16aligned_be32(&gnh->vni, htonl(ntohll(params->flow->tunnel.tun_id) << 8));
6b241d64
PS
702
703 ovs_mutex_unlock(&dev->mutex);
704
4975aa3e 705 opt_len = tun_metadata_to_geneve_header(&params->flow->tunnel,
6b241d64
PS
706 gnh->options, &crit_opt);
707
708 gnh->opt_len = opt_len / 4;
4975aa3e 709 gnh->oam = !!(params->flow->tunnel.flags & FLOW_TNL_F_OAM);
6b241d64
PS
710 gnh->critical = crit_opt ? 1 : 0;
711 gnh->proto_type = htons(ETH_TYPE_TEB);
712
4975aa3e 713 data->header_len += sizeof *gnh + opt_len;
6b241d64
PS
714 data->tnl_type = OVS_VPORT_TYPE_GENEVE;
715 return 0;
716}
717
718\f
719void
720netdev_tnl_egress_port_range(struct unixctl_conn *conn, int argc,
721 const char *argv[], void *aux OVS_UNUSED)
722{
723 int val1, val2;
724
725 if (argc < 3) {
726 struct ds ds = DS_EMPTY_INITIALIZER;
727
728 ds_put_format(&ds, "Tunnel UDP source port range: %"PRIu16"-%"PRIu16"\n",
729 tnl_udp_port_min, tnl_udp_port_max);
730
731 unixctl_command_reply(conn, ds_cstr(&ds));
732 ds_destroy(&ds);
733 return;
734 }
735
736 if (argc != 3) {
737 return;
738 }
739
740 val1 = atoi(argv[1]);
741 if (val1 <= 0 || val1 > UINT16_MAX) {
742 unixctl_command_reply(conn, "Invalid min.");
743 return;
744 }
745 val2 = atoi(argv[2]);
746 if (val2 <= 0 || val2 > UINT16_MAX) {
747 unixctl_command_reply(conn, "Invalid max.");
748 return;
749 }
750
751 if (val1 > val2) {
752 tnl_udp_port_min = val2;
753 tnl_udp_port_max = val1;
754 } else {
755 tnl_udp_port_min = val1;
756 tnl_udp_port_max = val2;
757 }
758 seq_change(tnl_conf_seq);
759
760 unixctl_command_reply(conn, "OK");
761}