1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 2010-2015 Intel Corporation
6 #include <linux/if_ether.h>
7 #include <linux/if_vlan.h>
8 #include <linux/virtio_net.h>
9 #include <linux/virtio_ring.h>
10 #include <sys/param.h>
13 #include <rte_ethdev.h>
15 #include <rte_string_fns.h>
17 #include <rte_malloc.h>
23 #include "rte_vhost.h"
25 #include "vxlan_setup.h"
27 #define IPV4_HEADER_LEN 20
28 #define UDP_HEADER_LEN 8
29 #define VXLAN_HEADER_LEN 8
31 #define IP_VERSION 0x40
32 #define IP_HDRLEN 0x05 /* default IP header length == five 32-bits words. */
33 #define IP_DEFTTL 64 /* from RFC 1340. */
34 #define IP_VHL_DEF (IP_VERSION | IP_HDRLEN)
36 #define IP_DN_FRAGMENT_FLAG 0x0040
38 /* Used to compare MAC addresses. */
39 #define MAC_ADDR_CMP 0xFFFFFFFFFFFFULL
41 /* Configurable number of RX/TX ring descriptors */
42 #define RTE_TEST_RX_DESC_DEFAULT 1024
43 #define RTE_TEST_TX_DESC_DEFAULT 512
45 /* Default inner VLAN ID */
46 #define INNER_VLAN_ID 100
49 struct vxlan_conf vxdev;
51 struct ipv4_hdr app_ip_hdr[VXLAN_N_PORTS];
52 struct ether_hdr app_l2_hdr[VXLAN_N_PORTS];
54 /* local VTEP IP address */
55 uint8_t vxlan_multicast_ips[2][4] = { {239, 1, 1, 1 }, {239, 1, 2, 1 } };
57 /* Remote VTEP IP address */
58 uint8_t vxlan_overlay_ips[2][4] = { {192, 168, 10, 1}, {192, 168, 30, 1} };
60 /* Remote VTEP MAC address */
61 uint8_t peer_mac[6] = {0x00, 0x11, 0x01, 0x00, 0x00, 0x01};
63 /* VXLAN RX filter type */
64 uint8_t tep_filter_type[] = {RTE_TUNNEL_FILTER_IMAC_TENID,
65 RTE_TUNNEL_FILTER_IMAC_IVLAN_TENID,
66 RTE_TUNNEL_FILTER_OMAC_TENID_IMAC,};
68 /* Options for configuring ethernet port */
69 static const struct rte_eth_conf port_conf = {
72 .header_split = 0, /**< Header Split disabled */
73 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
74 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
75 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
76 .hw_strip_crc = 1, /**< CRC stripped by hardware */
79 .mq_mode = ETH_MQ_TX_NONE,
84 * The one or two device(s) that belongs to the same tenant ID can
85 * be assigned in a VM.
87 const uint16_t tenant_id_conf[] = {
88 1000, 1000, 1001, 1001, 1002, 1002, 1003, 1003,
89 1004, 1004, 1005, 1005, 1006, 1006, 1007, 1007,
90 1008, 1008, 1009, 1009, 1010, 1010, 1011, 1011,
91 1012, 1012, 1013, 1013, 1014, 1014, 1015, 1015,
92 1016, 1016, 1017, 1017, 1018, 1018, 1019, 1019,
93 1020, 1020, 1021, 1021, 1022, 1022, 1023, 1023,
94 1024, 1024, 1025, 1025, 1026, 1026, 1027, 1027,
95 1028, 1028, 1029, 1029, 1030, 1030, 1031, 1031,
99 * Initialises a given port using global settings and with the rx buffers
100 * coming from the mbuf_pool passed as parameter
103 vxlan_port_init(uint16_t port, struct rte_mempool *mbuf_pool)
107 struct rte_eth_dev_info dev_info;
108 uint16_t rx_rings, tx_rings = (uint16_t)rte_lcore_count();
109 uint16_t rx_ring_size = RTE_TEST_RX_DESC_DEFAULT;
110 uint16_t tx_ring_size = RTE_TEST_TX_DESC_DEFAULT;
111 struct rte_eth_udp_tunnel tunnel_udp;
112 struct rte_eth_rxconf *rxconf;
113 struct rte_eth_txconf *txconf;
114 struct vxlan_conf *pconf = &vxdev;
116 pconf->dst_port = udp_port;
118 rte_eth_dev_info_get(port, &dev_info);
120 if (dev_info.max_rx_queues > MAX_QUEUES) {
121 rte_exit(EXIT_FAILURE,
122 "please define MAX_QUEUES no less than %u in %s\n",
123 dev_info.max_rx_queues, __FILE__);
126 rxconf = &dev_info.default_rxconf;
127 txconf = &dev_info.default_txconf;
128 txconf->txq_flags = 0;
130 if (port >= rte_eth_dev_count())
133 rx_rings = nb_devices;
135 /* Configure ethernet device. */
136 retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
140 retval = rte_eth_dev_adjust_nb_rx_tx_desc(port, &rx_ring_size,
145 /* Setup the queues. */
146 for (q = 0; q < rx_rings; q++) {
147 retval = rte_eth_rx_queue_setup(port, q, rx_ring_size,
148 rte_eth_dev_socket_id(port),
154 for (q = 0; q < tx_rings; q++) {
155 retval = rte_eth_tx_queue_setup(port, q, tx_ring_size,
156 rte_eth_dev_socket_id(port),
162 /* Start the device. */
163 retval = rte_eth_dev_start(port);
167 /* Configure UDP port for UDP tunneling */
168 tunnel_udp.udp_port = udp_port;
169 tunnel_udp.prot_type = RTE_TUNNEL_TYPE_VXLAN;
170 retval = rte_eth_dev_udp_tunnel_port_add(port, &tunnel_udp);
173 rte_eth_macaddr_get(port, &ports_eth_addr[port]);
174 RTE_LOG(INFO, PORT, "Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
175 " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
177 ports_eth_addr[port].addr_bytes[0],
178 ports_eth_addr[port].addr_bytes[1],
179 ports_eth_addr[port].addr_bytes[2],
180 ports_eth_addr[port].addr_bytes[3],
181 ports_eth_addr[port].addr_bytes[4],
182 ports_eth_addr[port].addr_bytes[5]);
184 if (tso_segsz != 0) {
185 struct rte_eth_dev_info dev_info;
186 rte_eth_dev_info_get(port, &dev_info);
187 if ((dev_info.tx_offload_capa & DEV_TX_OFFLOAD_TCP_TSO) == 0)
188 RTE_LOG(WARNING, PORT,
189 "hardware TSO offload is not supported\n");
195 vxlan_rx_process(struct rte_mbuf *pkt)
200 ret = decapsulation(pkt);
206 vxlan_tx_process(uint8_t queue_id, struct rte_mbuf *pkt)
209 encapsulation(pkt, queue_id);
215 * This function learns the MAC address of the device and set init
216 * L2 header and L3 header info.
219 vxlan_link(struct vhost_dev *vdev, struct rte_mbuf *m)
222 struct ether_hdr *pkt_hdr;
223 uint64_t portid = vdev->vid;
226 struct rte_eth_tunnel_filter_conf tunnel_filter_conf;
228 if (unlikely(portid >= VXLAN_N_PORTS)) {
229 RTE_LOG(INFO, VHOST_DATA,
230 "(%d) WARNING: Not configuring device,"
231 "as already have %d ports for VXLAN.",
232 vdev->vid, VXLAN_N_PORTS);
236 /* Learn MAC address of guest device from packet */
237 pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
238 if (is_same_ether_addr(&(pkt_hdr->s_addr), &vdev->mac_address)) {
239 RTE_LOG(INFO, VHOST_DATA,
240 "(%d) WARNING: This device is using an existing"
241 " MAC address and has not been registered.\n",
246 for (i = 0; i < ETHER_ADDR_LEN; i++) {
247 vdev->mac_address.addr_bytes[i] =
248 vxdev.port[portid].vport_mac.addr_bytes[i] =
249 pkt_hdr->s_addr.addr_bytes[i];
250 vxdev.port[portid].peer_mac.addr_bytes[i] = peer_mac[i];
253 memset(&tunnel_filter_conf, 0,
254 sizeof(struct rte_eth_tunnel_filter_conf));
256 ether_addr_copy(&ports_eth_addr[0], &tunnel_filter_conf.outer_mac);
257 tunnel_filter_conf.filter_type = tep_filter_type[filter_idx];
260 ether_addr_copy(&vdev->mac_address, &tunnel_filter_conf.inner_mac);
262 tunnel_filter_conf.queue_id = vdev->rx_q;
263 tunnel_filter_conf.tenant_id = tenant_id_conf[vdev->rx_q];
265 if (tep_filter_type[filter_idx] == RTE_TUNNEL_FILTER_IMAC_IVLAN_TENID)
266 tunnel_filter_conf.inner_vlan = INNER_VLAN_ID;
268 tunnel_filter_conf.tunnel_type = RTE_TUNNEL_TYPE_VXLAN;
270 ret = rte_eth_dev_filter_ctrl(ports[0],
271 RTE_ETH_FILTER_TUNNEL,
273 &tunnel_filter_conf);
275 RTE_LOG(ERR, VHOST_DATA,
276 "%d Failed to add device MAC address to cloud filter\n",
281 /* Print out inner MAC and VNI info. */
282 RTE_LOG(INFO, VHOST_DATA,
283 "(%d) MAC_ADDRESS %02x:%02x:%02x:%02x:%02x:%02x and VNI %d registered\n",
285 vdev->mac_address.addr_bytes[0],
286 vdev->mac_address.addr_bytes[1],
287 vdev->mac_address.addr_bytes[2],
288 vdev->mac_address.addr_bytes[3],
289 vdev->mac_address.addr_bytes[4],
290 vdev->mac_address.addr_bytes[5],
291 tenant_id_conf[vdev->rx_q]);
293 vxdev.port[portid].vport_id = portid;
295 for (i = 0; i < 4; i++) {
297 vxdev.port_ip |= vxlan_multicast_ips[portid][i] << (8 * i);
299 vxdev.port[portid].peer_ip |=
300 vxlan_overlay_ips[portid][i] << (8 * i);
303 vxdev.out_key = tenant_id_conf[vdev->rx_q];
304 ether_addr_copy(&vxdev.port[portid].peer_mac,
305 &app_l2_hdr[portid].d_addr);
306 ether_addr_copy(&ports_eth_addr[0],
307 &app_l2_hdr[portid].s_addr);
308 app_l2_hdr[portid].ether_type = rte_cpu_to_be_16(ETHER_TYPE_IPv4);
310 ip = &app_ip_hdr[portid];
311 ip->version_ihl = IP_VHL_DEF;
312 ip->type_of_service = 0;
313 ip->total_length = 0;
315 ip->fragment_offset = IP_DN_FRAGMENT_FLAG;
316 ip->time_to_live = IP_DEFTTL;
317 ip->next_proto_id = IPPROTO_UDP;
318 ip->hdr_checksum = 0;
319 ip->src_addr = vxdev.port_ip;
320 ip->dst_addr = vxdev.port[portid].peer_ip;
322 /* Set device as ready for RX. */
323 vdev->ready = DEVICE_RX;
329 * Removes cloud filter. Ensures that nothing is adding buffers to the RX
330 * queue before disabling RX on the device.
333 vxlan_unlink(struct vhost_dev *vdev)
335 unsigned i = 0, rx_count;
337 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
338 struct rte_eth_tunnel_filter_conf tunnel_filter_conf;
340 if (vdev->ready == DEVICE_RX) {
341 memset(&tunnel_filter_conf, 0,
342 sizeof(struct rte_eth_tunnel_filter_conf));
344 ether_addr_copy(&ports_eth_addr[0], &tunnel_filter_conf.outer_mac);
345 ether_addr_copy(&vdev->mac_address, &tunnel_filter_conf.inner_mac);
346 tunnel_filter_conf.tenant_id = tenant_id_conf[vdev->rx_q];
347 tunnel_filter_conf.filter_type = tep_filter_type[filter_idx];
349 if (tep_filter_type[filter_idx] ==
350 RTE_TUNNEL_FILTER_IMAC_IVLAN_TENID)
351 tunnel_filter_conf.inner_vlan = INNER_VLAN_ID;
353 tunnel_filter_conf.queue_id = vdev->rx_q;
354 tunnel_filter_conf.tunnel_type = RTE_TUNNEL_TYPE_VXLAN;
356 ret = rte_eth_dev_filter_ctrl(ports[0],
357 RTE_ETH_FILTER_TUNNEL,
358 RTE_ETH_FILTER_DELETE,
359 &tunnel_filter_conf);
361 RTE_LOG(ERR, VHOST_DATA,
362 "%d Failed to add device MAC address to cloud filter\n",
366 for (i = 0; i < ETHER_ADDR_LEN; i++)
367 vdev->mac_address.addr_bytes[i] = 0;
369 /* Clear out the receive buffers */
370 rx_count = rte_eth_rx_burst(ports[0],
371 (uint16_t)vdev->rx_q,
372 pkts_burst, MAX_PKT_BURST);
375 for (i = 0; i < rx_count; i++)
376 rte_pktmbuf_free(pkts_burst[i]);
378 rx_count = rte_eth_rx_burst(ports[0],
379 (uint16_t)vdev->rx_q,
380 pkts_burst, MAX_PKT_BURST);
382 vdev->ready = DEVICE_MAC_LEARNING;
386 /* Transmit packets after encapsulating */
388 vxlan_tx_pkts(uint16_t port_id, uint16_t queue_id,
389 struct rte_mbuf **tx_pkts, uint16_t nb_pkts) {
393 for (i = 0; i < nb_pkts; i++)
394 vxlan_tx_process(queue_id, tx_pkts[i]);
396 ret = rte_eth_tx_burst(port_id, queue_id, tx_pkts, nb_pkts);
401 /* Check for decapsulation and pass packets directly to VIRTIO device */
403 vxlan_rx_pkts(int vid, struct rte_mbuf **pkts_burst, uint32_t rx_count)
408 struct rte_mbuf *pkts_valid[rx_count];
410 for (i = 0; i < rx_count; i++) {
413 &dev_statistics[vid].rx_bad_ip_csum,
414 (pkts_burst[i]->ol_flags & PKT_RX_IP_CKSUM_BAD)
417 &dev_statistics[vid].rx_bad_ip_csum,
418 (pkts_burst[i]->ol_flags & PKT_RX_L4_CKSUM_BAD)
421 ret = vxlan_rx_process(pkts_burst[i]);
422 if (unlikely(ret < 0))
425 pkts_valid[count] = pkts_burst[i];
429 ret = rte_vhost_enqueue_burst(vid, VIRTIO_RXQ, pkts_valid, count);