4 * Copyright(c) 2010-2016 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 #include <arpa/inet.h>
36 #include <linux/if_ether.h>
37 #include <linux/if_vlan.h>
38 #include <linux/virtio_net.h>
39 #include <linux/virtio_ring.h>
42 #include <sys/eventfd.h>
43 #include <sys/param.h>
46 #include <rte_atomic.h>
47 #include <rte_cycles.h>
48 #include <rte_ethdev.h>
50 #include <rte_string_fns.h>
51 #include <rte_malloc.h>
52 #include <rte_virtio_net.h>
59 #define MAX_QUEUES 128
62 /* the maximum number of external ports supported */
63 #define MAX_SUP_PORTS 1
65 #define MBUF_CACHE_SIZE 128
66 #define MBUF_DATA_SIZE RTE_MBUF_DEFAULT_BUF_SIZE
68 #define MAX_PKT_BURST 32 /* Max burst size for RX/TX */
69 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
71 #define BURST_RX_WAIT_US 15 /* Defines how long we wait between retries on RX */
72 #define BURST_RX_RETRIES 4 /* Number of retries on RX. */
74 #define JUMBO_FRAME_MAX_SIZE 0x2600
76 /* State of virtio device. */
77 #define DEVICE_MAC_LEARNING 0
79 #define DEVICE_SAFE_REMOVE 2
81 /* Configurable number of RX/TX ring descriptors */
82 #define RTE_TEST_RX_DESC_DEFAULT 1024
83 #define RTE_TEST_TX_DESC_DEFAULT 512
85 #define INVALID_PORT_ID 0xFF
87 /* Max number of devices. Limited by vmdq. */
88 #define MAX_DEVICES 64
90 /* Size of buffers used for snprintfs. */
91 #define MAX_PRINT_BUFF 6072
93 /* Maximum character device basename size. */
94 #define MAX_BASENAME_SZ 10
96 /* Maximum long option length for option parsing. */
97 #define MAX_LONG_OPT_SZ 64
99 /* mask of enabled ports */
100 static uint32_t enabled_port_mask = 0;
102 /* Promiscuous mode */
103 static uint32_t promiscuous;
105 /* number of devices/queues to support*/
106 static uint32_t num_queues = 0;
107 static uint32_t num_devices;
109 static struct rte_mempool *mbuf_pool;
110 static int mergeable;
112 /* Do vlan strip on host, enabled on default */
113 static uint32_t vlan_strip = 1;
115 /* Enable VM2VM communications. If this is disabled then the MAC address compare is skipped. */
122 static vm2vm_type vm2vm_mode = VM2VM_SOFTWARE;
125 static uint32_t enable_stats = 0;
126 /* Enable retries on RX. */
127 static uint32_t enable_retry = 1;
129 /* Disable TX checksum offload */
130 static uint32_t enable_tx_csum;
132 /* Disable TSO offload */
133 static uint32_t enable_tso;
135 /* Specify timeout (in useconds) between retries on RX. */
136 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
137 /* Specify the number of retries on RX. */
138 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
140 /* Character device basename. Can be set by user. */
141 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
143 /* empty vmdq configuration structure. Filled in programatically */
144 static struct rte_eth_conf vmdq_conf_default = {
146 .mq_mode = ETH_MQ_RX_VMDQ_ONLY,
148 .header_split = 0, /**< Header Split disabled */
149 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
150 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
152 * It is necessary for 1G NIC such as I350,
153 * this fixes bug of ipv4 forwarding in guest can't
154 * forward pakets from one virtio dev to another virtio dev.
156 .hw_vlan_strip = 1, /**< VLAN strip enabled. */
157 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
158 .hw_strip_crc = 0, /**< CRC stripped by hardware */
162 .mq_mode = ETH_MQ_TX_NONE,
166 * should be overridden separately in code with
170 .nb_queue_pools = ETH_8_POOLS,
171 .enable_default_pool = 0,
174 .pool_map = {{0, 0},},
179 static unsigned lcore_ids[RTE_MAX_LCORE];
180 static uint8_t ports[RTE_MAX_ETHPORTS];
181 static unsigned num_ports = 0; /**< The number of ports specified in command line */
182 static uint16_t num_pf_queues, num_vmdq_queues;
183 static uint16_t vmdq_pool_base, vmdq_queue_base;
184 static uint16_t queues_per_pool;
186 const uint16_t vlan_tags[] = {
187 1000, 1001, 1002, 1003, 1004, 1005, 1006, 1007,
188 1008, 1009, 1010, 1011, 1012, 1013, 1014, 1015,
189 1016, 1017, 1018, 1019, 1020, 1021, 1022, 1023,
190 1024, 1025, 1026, 1027, 1028, 1029, 1030, 1031,
191 1032, 1033, 1034, 1035, 1036, 1037, 1038, 1039,
192 1040, 1041, 1042, 1043, 1044, 1045, 1046, 1047,
193 1048, 1049, 1050, 1051, 1052, 1053, 1054, 1055,
194 1056, 1057, 1058, 1059, 1060, 1061, 1062, 1063,
197 /* ethernet addresses of ports */
198 static struct ether_addr vmdq_ports_eth_addr[RTE_MAX_ETHPORTS];
200 static struct vhost_dev_tailq_list vhost_dev_list =
201 TAILQ_HEAD_INITIALIZER(vhost_dev_list);
203 static struct lcore_info lcore_info[RTE_MAX_LCORE];
205 /* Used for queueing bursts of TX packets. */
209 struct rte_mbuf *m_table[MAX_PKT_BURST];
212 /* TX queue for each data core. */
213 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
215 #define MBUF_TABLE_DRAIN_TSC ((rte_get_tsc_hz() + US_PER_S - 1) \
216 / US_PER_S * BURST_TX_DRAIN_US)
219 /* Per-device statistics struct */
220 struct device_statistics {
222 rte_atomic64_t rx_total_atomic;
224 rte_atomic64_t rx_atomic;
225 } __rte_cache_aligned;
226 struct device_statistics dev_statistics[MAX_DEVICES];
229 * Builds up the correct configuration for VMDQ VLAN pool map
230 * according to the pool & queue limits.
233 get_eth_conf(struct rte_eth_conf *eth_conf, uint32_t num_devices)
235 struct rte_eth_vmdq_rx_conf conf;
236 struct rte_eth_vmdq_rx_conf *def_conf =
237 &vmdq_conf_default.rx_adv_conf.vmdq_rx_conf;
240 memset(&conf, 0, sizeof(conf));
241 conf.nb_queue_pools = (enum rte_eth_nb_pools)num_devices;
242 conf.nb_pool_maps = num_devices;
243 conf.enable_loop_back = def_conf->enable_loop_back;
244 conf.rx_mode = def_conf->rx_mode;
246 for (i = 0; i < conf.nb_pool_maps; i++) {
247 conf.pool_map[i].vlan_id = vlan_tags[ i ];
248 conf.pool_map[i].pools = (1UL << i);
251 (void)(rte_memcpy(eth_conf, &vmdq_conf_default, sizeof(*eth_conf)));
252 (void)(rte_memcpy(ð_conf->rx_adv_conf.vmdq_rx_conf, &conf,
253 sizeof(eth_conf->rx_adv_conf.vmdq_rx_conf)));
258 * Validate the device number according to the max pool number gotten form
259 * dev_info. If the device number is invalid, give the error message and
260 * return -1. Each device must have its own pool.
263 validate_num_devices(uint32_t max_nb_devices)
265 if (num_devices > max_nb_devices) {
266 RTE_LOG(ERR, VHOST_PORT, "invalid number of devices\n");
273 * Initialises a given port using global settings and with the rx buffers
274 * coming from the mbuf_pool passed as parameter
277 port_init(uint8_t port)
279 struct rte_eth_dev_info dev_info;
280 struct rte_eth_conf port_conf;
281 struct rte_eth_rxconf *rxconf;
282 struct rte_eth_txconf *txconf;
283 int16_t rx_rings, tx_rings;
284 uint16_t rx_ring_size, tx_ring_size;
288 /* The max pool number from dev_info will be used to validate the pool number specified in cmd line */
289 rte_eth_dev_info_get (port, &dev_info);
291 if (dev_info.max_rx_queues > MAX_QUEUES) {
292 rte_exit(EXIT_FAILURE,
293 "please define MAX_QUEUES no less than %u in %s\n",
294 dev_info.max_rx_queues, __FILE__);
297 rxconf = &dev_info.default_rxconf;
298 txconf = &dev_info.default_txconf;
299 rxconf->rx_drop_en = 1;
301 /* Enable vlan offload */
302 txconf->txq_flags &= ~ETH_TXQ_FLAGS_NOVLANOFFL;
304 /*configure the number of supported virtio devices based on VMDQ limits */
305 num_devices = dev_info.max_vmdq_pools;
307 rx_ring_size = RTE_TEST_RX_DESC_DEFAULT;
308 tx_ring_size = RTE_TEST_TX_DESC_DEFAULT;
309 tx_rings = (uint16_t)rte_lcore_count();
311 retval = validate_num_devices(MAX_DEVICES);
315 /* Get port configuration. */
316 retval = get_eth_conf(&port_conf, num_devices);
319 /* NIC queues are divided into pf queues and vmdq queues. */
320 num_pf_queues = dev_info.max_rx_queues - dev_info.vmdq_queue_num;
321 queues_per_pool = dev_info.vmdq_queue_num / dev_info.max_vmdq_pools;
322 num_vmdq_queues = num_devices * queues_per_pool;
323 num_queues = num_pf_queues + num_vmdq_queues;
324 vmdq_queue_base = dev_info.vmdq_queue_base;
325 vmdq_pool_base = dev_info.vmdq_pool_base;
326 printf("pf queue num: %u, configured vmdq pool num: %u, each vmdq pool has %u queues\n",
327 num_pf_queues, num_devices, queues_per_pool);
329 if (port >= rte_eth_dev_count()) return -1;
331 if (enable_tx_csum == 0)
332 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_CSUM);
334 if (enable_tso == 0) {
335 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO4);
336 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_HOST_TSO6);
339 rx_rings = (uint16_t)dev_info.max_rx_queues;
340 /* Configure ethernet device. */
341 retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
345 /* Setup the queues. */
346 for (q = 0; q < rx_rings; q ++) {
347 retval = rte_eth_rx_queue_setup(port, q, rx_ring_size,
348 rte_eth_dev_socket_id(port),
354 for (q = 0; q < tx_rings; q ++) {
355 retval = rte_eth_tx_queue_setup(port, q, tx_ring_size,
356 rte_eth_dev_socket_id(port),
362 /* Start the device. */
363 retval = rte_eth_dev_start(port);
365 RTE_LOG(ERR, VHOST_DATA, "Failed to start the device.\n");
370 rte_eth_promiscuous_enable(port);
372 rte_eth_macaddr_get(port, &vmdq_ports_eth_addr[port]);
373 RTE_LOG(INFO, VHOST_PORT, "Max virtio devices supported: %u\n", num_devices);
374 RTE_LOG(INFO, VHOST_PORT, "Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
375 " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
377 vmdq_ports_eth_addr[port].addr_bytes[0],
378 vmdq_ports_eth_addr[port].addr_bytes[1],
379 vmdq_ports_eth_addr[port].addr_bytes[2],
380 vmdq_ports_eth_addr[port].addr_bytes[3],
381 vmdq_ports_eth_addr[port].addr_bytes[4],
382 vmdq_ports_eth_addr[port].addr_bytes[5]);
388 * Set character device basename.
391 us_vhost_parse_basename(const char *q_arg)
393 /* parse number string */
395 if (strnlen(q_arg, MAX_BASENAME_SZ) > MAX_BASENAME_SZ)
398 snprintf((char*)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
404 * Parse the portmask provided at run time.
407 parse_portmask(const char *portmask)
414 /* parse hexadecimal string */
415 pm = strtoul(portmask, &end, 16);
416 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
427 * Parse num options at run time.
430 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
437 /* parse unsigned int string */
438 num = strtoul(q_arg, &end, 10);
439 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
442 if (num > max_valid_value)
453 us_vhost_usage(const char *prgname)
455 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
457 " --rx_retry [0|1] --mergeable [0|1] --stats [0-N]\n"
458 " --dev-basename <name>\n"
460 " -p PORTMASK: Set mask for ports to be used by application\n"
461 " --vm2vm [0|1|2]: disable/software(default)/hardware vm2vm comms\n"
462 " --rx-retry [0|1]: disable/enable(default) retries on rx. Enable retry if destintation queue is full\n"
463 " --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX. This makes effect only if retries on rx enabled\n"
464 " --rx-retry-num [0-N]: the number of retries on rx. This makes effect only if retries on rx enabled\n"
465 " --mergeable [0|1]: disable(default)/enable RX mergeable buffers\n"
466 " --vlan-strip [0|1]: disable/enable(default) RX VLAN strip on host\n"
467 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
468 " --dev-basename: The basename to be used for the character device.\n"
469 " --tx-csum [0|1] disable/enable TX checksum offload.\n"
470 " --tso [0|1] disable/enable TCP segment offload.\n",
475 * Parse the arguments given in the command line of the application.
478 us_vhost_parse_args(int argc, char **argv)
483 const char *prgname = argv[0];
484 static struct option long_option[] = {
485 {"vm2vm", required_argument, NULL, 0},
486 {"rx-retry", required_argument, NULL, 0},
487 {"rx-retry-delay", required_argument, NULL, 0},
488 {"rx-retry-num", required_argument, NULL, 0},
489 {"mergeable", required_argument, NULL, 0},
490 {"vlan-strip", required_argument, NULL, 0},
491 {"stats", required_argument, NULL, 0},
492 {"dev-basename", required_argument, NULL, 0},
493 {"tx-csum", required_argument, NULL, 0},
494 {"tso", required_argument, NULL, 0},
498 /* Parse command line */
499 while ((opt = getopt_long(argc, argv, "p:P",
500 long_option, &option_index)) != EOF) {
504 enabled_port_mask = parse_portmask(optarg);
505 if (enabled_port_mask == 0) {
506 RTE_LOG(INFO, VHOST_CONFIG, "Invalid portmask\n");
507 us_vhost_usage(prgname);
514 vmdq_conf_default.rx_adv_conf.vmdq_rx_conf.rx_mode =
515 ETH_VMDQ_ACCEPT_BROADCAST |
516 ETH_VMDQ_ACCEPT_MULTICAST;
517 rte_vhost_feature_enable(1ULL << VIRTIO_NET_F_CTRL_RX);
522 /* Enable/disable vm2vm comms. */
523 if (!strncmp(long_option[option_index].name, "vm2vm",
525 ret = parse_num_opt(optarg, (VM2VM_LAST - 1));
527 RTE_LOG(INFO, VHOST_CONFIG,
528 "Invalid argument for "
530 us_vhost_usage(prgname);
533 vm2vm_mode = (vm2vm_type)ret;
537 /* Enable/disable retries on RX. */
538 if (!strncmp(long_option[option_index].name, "rx-retry", MAX_LONG_OPT_SZ)) {
539 ret = parse_num_opt(optarg, 1);
541 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for rx-retry [0|1]\n");
542 us_vhost_usage(prgname);
549 /* Enable/disable TX checksum offload. */
550 if (!strncmp(long_option[option_index].name, "tx-csum", MAX_LONG_OPT_SZ)) {
551 ret = parse_num_opt(optarg, 1);
553 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for tx-csum [0|1]\n");
554 us_vhost_usage(prgname);
557 enable_tx_csum = ret;
560 /* Enable/disable TSO offload. */
561 if (!strncmp(long_option[option_index].name, "tso", MAX_LONG_OPT_SZ)) {
562 ret = parse_num_opt(optarg, 1);
564 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for tso [0|1]\n");
565 us_vhost_usage(prgname);
571 /* Specify the retries delay time (in useconds) on RX. */
572 if (!strncmp(long_option[option_index].name, "rx-retry-delay", MAX_LONG_OPT_SZ)) {
573 ret = parse_num_opt(optarg, INT32_MAX);
575 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for rx-retry-delay [0-N]\n");
576 us_vhost_usage(prgname);
579 burst_rx_delay_time = ret;
583 /* Specify the retries number on RX. */
584 if (!strncmp(long_option[option_index].name, "rx-retry-num", MAX_LONG_OPT_SZ)) {
585 ret = parse_num_opt(optarg, INT32_MAX);
587 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for rx-retry-num [0-N]\n");
588 us_vhost_usage(prgname);
591 burst_rx_retry_num = ret;
595 /* Enable/disable RX mergeable buffers. */
596 if (!strncmp(long_option[option_index].name, "mergeable", MAX_LONG_OPT_SZ)) {
597 ret = parse_num_opt(optarg, 1);
599 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for mergeable [0|1]\n");
600 us_vhost_usage(prgname);
605 vmdq_conf_default.rxmode.jumbo_frame = 1;
606 vmdq_conf_default.rxmode.max_rx_pkt_len
607 = JUMBO_FRAME_MAX_SIZE;
612 /* Enable/disable RX VLAN strip on host. */
613 if (!strncmp(long_option[option_index].name,
614 "vlan-strip", MAX_LONG_OPT_SZ)) {
615 ret = parse_num_opt(optarg, 1);
617 RTE_LOG(INFO, VHOST_CONFIG,
618 "Invalid argument for VLAN strip [0|1]\n");
619 us_vhost_usage(prgname);
623 vmdq_conf_default.rxmode.hw_vlan_strip =
628 /* Enable/disable stats. */
629 if (!strncmp(long_option[option_index].name, "stats", MAX_LONG_OPT_SZ)) {
630 ret = parse_num_opt(optarg, INT32_MAX);
632 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for stats [0..N]\n");
633 us_vhost_usage(prgname);
640 /* Set character device basename. */
641 if (!strncmp(long_option[option_index].name, "dev-basename", MAX_LONG_OPT_SZ)) {
642 if (us_vhost_parse_basename(optarg) == -1) {
643 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for character device basename (Max %d characters)\n", MAX_BASENAME_SZ);
644 us_vhost_usage(prgname);
651 /* Invalid option - print options. */
653 us_vhost_usage(prgname);
658 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
659 if (enabled_port_mask & (1 << i))
660 ports[num_ports++] = (uint8_t)i;
663 if ((num_ports == 0) || (num_ports > MAX_SUP_PORTS)) {
664 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
665 "but only %u port can be enabled\n",num_ports, MAX_SUP_PORTS);
673 * Update the global var NUM_PORTS and array PORTS according to system ports number
674 * and return valid ports number
676 static unsigned check_ports_num(unsigned nb_ports)
678 unsigned valid_num_ports = num_ports;
681 if (num_ports > nb_ports) {
682 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) exceeds total system port number(%u)\n",
683 num_ports, nb_ports);
684 num_ports = nb_ports;
687 for (portid = 0; portid < num_ports; portid ++) {
688 if (ports[portid] >= nb_ports) {
689 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port ID(%u) exceeds max system port ID(%u)\n",
690 ports[portid], (nb_ports - 1));
691 ports[portid] = INVALID_PORT_ID;
695 return valid_num_ports;
698 static inline struct vhost_dev *__attribute__((always_inline))
699 find_vhost_dev(struct ether_addr *mac)
701 struct vhost_dev *vdev;
703 TAILQ_FOREACH(vdev, &vhost_dev_list, next) {
704 if (vdev->ready == DEVICE_RX &&
705 is_same_ether_addr(mac, &vdev->mac_address))
713 * This function learns the MAC address of the device and registers this along with a
714 * vlan tag to a VMDQ.
717 link_vmdq(struct vhost_dev *vdev, struct rte_mbuf *m)
719 struct ether_hdr *pkt_hdr;
720 struct virtio_net *dev = vdev->dev;
723 /* Learn MAC address of guest device from packet */
724 pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
726 if (find_vhost_dev(&pkt_hdr->s_addr)) {
727 RTE_LOG(ERR, VHOST_DATA,
728 "Device (%" PRIu64 ") is using a registered MAC!\n",
733 for (i = 0; i < ETHER_ADDR_LEN; i++)
734 vdev->mac_address.addr_bytes[i] = pkt_hdr->s_addr.addr_bytes[i];
736 /* vlan_tag currently uses the device_id. */
737 vdev->vlan_tag = vlan_tags[dev->device_fh];
739 /* Print out VMDQ registration info. */
740 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") MAC_ADDRESS %02x:%02x:%02x:%02x:%02x:%02x and VLAN_TAG %d registered\n",
742 vdev->mac_address.addr_bytes[0], vdev->mac_address.addr_bytes[1],
743 vdev->mac_address.addr_bytes[2], vdev->mac_address.addr_bytes[3],
744 vdev->mac_address.addr_bytes[4], vdev->mac_address.addr_bytes[5],
747 /* Register the MAC address. */
748 ret = rte_eth_dev_mac_addr_add(ports[0], &vdev->mac_address,
749 (uint32_t)dev->device_fh + vmdq_pool_base);
751 RTE_LOG(ERR, VHOST_DATA, "(%"PRIu64") Failed to add device MAC address to VMDQ\n",
754 /* Enable stripping of the vlan tag as we handle routing. */
756 rte_eth_dev_set_vlan_strip_on_queue(ports[0],
757 (uint16_t)vdev->vmdq_rx_q, 1);
759 /* Set device as ready for RX. */
760 vdev->ready = DEVICE_RX;
766 * Removes MAC address and vlan tag from VMDQ. Ensures that nothing is adding buffers to the RX
767 * queue before disabling RX on the device.
770 unlink_vmdq(struct vhost_dev *vdev)
774 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
776 if (vdev->ready == DEVICE_RX) {
777 /*clear MAC and VLAN settings*/
778 rte_eth_dev_mac_addr_remove(ports[0], &vdev->mac_address);
779 for (i = 0; i < 6; i++)
780 vdev->mac_address.addr_bytes[i] = 0;
784 /*Clear out the receive buffers*/
785 rx_count = rte_eth_rx_burst(ports[0],
786 (uint16_t)vdev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
789 for (i = 0; i < rx_count; i++)
790 rte_pktmbuf_free(pkts_burst[i]);
792 rx_count = rte_eth_rx_burst(ports[0],
793 (uint16_t)vdev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
796 vdev->ready = DEVICE_MAC_LEARNING;
800 static inline void __attribute__((always_inline))
801 virtio_xmit(struct virtio_net *dst_dev, struct virtio_net *src_dev,
806 ret = rte_vhost_enqueue_burst(dst_dev, VIRTIO_RXQ, &m, 1);
808 rte_atomic64_inc(&dev_statistics[dst_dev->device_fh].rx_total_atomic);
809 rte_atomic64_add(&dev_statistics[dst_dev->device_fh].rx_atomic, ret);
810 dev_statistics[src_dev->device_fh].tx_total++;
811 dev_statistics[src_dev->device_fh].tx += ret;
816 * Check if the packet destination MAC address is for a local device. If so then put
817 * the packet on that devices RX queue. If not then return.
819 static inline int __attribute__((always_inline))
820 virtio_tx_local(struct vhost_dev *vdev, struct rte_mbuf *m)
822 struct ether_hdr *pkt_hdr;
823 struct vhost_dev *dst_vdev;
826 pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
828 dst_vdev = find_vhost_dev(&pkt_hdr->d_addr);
832 fh = dst_vdev->dev->device_fh;
833 if (fh == vdev->dev->device_fh) {
834 RTE_LOG(DEBUG, VHOST_DATA,
835 "(%" PRIu64 ") TX: src and dst MAC is same. "
836 "Dropping packet.\n", fh);
840 RTE_LOG(DEBUG, VHOST_DATA,
841 "(%" PRIu64 ") TX: MAC address is local\n", fh);
843 if (unlikely(dst_vdev->remove)) {
844 RTE_LOG(DEBUG, VHOST_DATA, "(%" PRIu64 ") "
845 "Device is marked for removal\n", fh);
849 virtio_xmit(dst_vdev->dev, vdev->dev, m);
854 * Check if the destination MAC of a packet is one local VM,
855 * and get its vlan tag, and offset if it is.
857 static inline int __attribute__((always_inline))
858 find_local_dest(struct virtio_net *dev, struct rte_mbuf *m,
859 uint32_t *offset, uint16_t *vlan_tag)
861 struct vhost_dev *dst_vdev;
862 struct ether_hdr *pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
864 dst_vdev = find_vhost_dev(&pkt_hdr->d_addr);
868 if (dst_vdev->dev->device_fh == dev->device_fh) {
869 RTE_LOG(DEBUG, VHOST_DATA,
870 "(%" PRIu64 ") TX: src and dst MAC is same. "
871 " Dropping packet.\n", dst_vdev->dev->device_fh);
876 * HW vlan strip will reduce the packet length
877 * by minus length of vlan tag, so need restore
878 * the packet length by plus it.
881 *vlan_tag = vlan_tags[(uint16_t)dst_vdev->dev->device_fh];
883 RTE_LOG(DEBUG, VHOST_DATA,
884 "(%" PRIu64 ") TX: pkt to local VM device id: (%" PRIu64 ") "
886 dev->device_fh, dst_vdev->dev->device_fh, *vlan_tag);
892 get_psd_sum(void *l3_hdr, uint64_t ol_flags)
894 if (ol_flags & PKT_TX_IPV4)
895 return rte_ipv4_phdr_cksum(l3_hdr, ol_flags);
896 else /* assume ethertype == ETHER_TYPE_IPv6 */
897 return rte_ipv6_phdr_cksum(l3_hdr, ol_flags);
900 static void virtio_tx_offload(struct rte_mbuf *m)
903 struct ipv4_hdr *ipv4_hdr = NULL;
904 struct tcp_hdr *tcp_hdr = NULL;
905 struct ether_hdr *eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
907 l3_hdr = (char *)eth_hdr + m->l2_len;
909 if (m->ol_flags & PKT_TX_IPV4) {
911 ipv4_hdr->hdr_checksum = 0;
912 m->ol_flags |= PKT_TX_IP_CKSUM;
915 tcp_hdr = (struct tcp_hdr *)((char *)l3_hdr + m->l3_len);
916 tcp_hdr->cksum = get_psd_sum(l3_hdr, m->ol_flags);
920 free_pkts(struct rte_mbuf **pkts, uint16_t n)
923 rte_pktmbuf_free(pkts[n]);
926 static inline void __attribute__((always_inline))
927 do_drain_mbuf_table(struct mbuf_table *tx_q)
931 count = rte_eth_tx_burst(ports[0], tx_q->txq_id,
932 tx_q->m_table, tx_q->len);
933 if (unlikely(count < tx_q->len))
934 free_pkts(&tx_q->m_table[count], tx_q->len - count);
940 * This function routes the TX packet to the correct interface. This
941 * may be a local device or the physical port.
943 static inline void __attribute__((always_inline))
944 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
946 struct mbuf_table *tx_q;
948 const uint16_t lcore_id = rte_lcore_id();
949 struct virtio_net *dev = vdev->dev;
950 struct ether_hdr *nh;
953 nh = rte_pktmbuf_mtod(m, struct ether_hdr *);
954 if (unlikely(is_broadcast_ether_addr(&nh->d_addr))) {
955 struct vhost_dev *vdev2;
957 TAILQ_FOREACH(vdev2, &vhost_dev_list, next) {
958 virtio_xmit(vdev2->dev, vdev->dev, m);
963 /*check if destination is local VM*/
964 if ((vm2vm_mode == VM2VM_SOFTWARE) && (virtio_tx_local(vdev, m) == 0)) {
969 if (unlikely(vm2vm_mode == VM2VM_HARDWARE)) {
970 if (unlikely(find_local_dest(dev, m, &offset, &vlan_tag) != 0)) {
976 RTE_LOG(DEBUG, VHOST_DATA, "(%" PRIu64 ") TX: "
977 "MAC address is external\n", dev->device_fh);
981 /*Add packet to the port tx queue*/
982 tx_q = &lcore_tx_queue[lcore_id];
984 nh = rte_pktmbuf_mtod(m, struct ether_hdr *);
985 if (unlikely(nh->ether_type == rte_cpu_to_be_16(ETHER_TYPE_VLAN))) {
986 /* Guest has inserted the vlan tag. */
987 struct vlan_hdr *vh = (struct vlan_hdr *) (nh + 1);
988 uint16_t vlan_tag_be = rte_cpu_to_be_16(vlan_tag);
989 if ((vm2vm_mode == VM2VM_HARDWARE) &&
990 (vh->vlan_tci != vlan_tag_be))
991 vh->vlan_tci = vlan_tag_be;
993 m->ol_flags |= PKT_TX_VLAN_PKT;
996 * Find the right seg to adjust the data len when offset is
997 * bigger than tail room size.
999 if (unlikely(vm2vm_mode == VM2VM_HARDWARE)) {
1000 if (likely(offset <= rte_pktmbuf_tailroom(m)))
1001 m->data_len += offset;
1003 struct rte_mbuf *seg = m;
1005 while ((seg->next != NULL) &&
1006 (offset > rte_pktmbuf_tailroom(seg)))
1009 seg->data_len += offset;
1011 m->pkt_len += offset;
1014 m->vlan_tci = vlan_tag;
1017 if (m->ol_flags & PKT_TX_TCP_SEG)
1018 virtio_tx_offload(m);
1020 tx_q->m_table[tx_q->len++] = m;
1022 dev_statistics[dev->device_fh].tx_total++;
1023 dev_statistics[dev->device_fh].tx++;
1026 if (unlikely(tx_q->len == MAX_PKT_BURST))
1027 do_drain_mbuf_table(tx_q);
1031 static inline void __attribute__((always_inline))
1032 drain_mbuf_table(struct mbuf_table *tx_q)
1034 static uint64_t prev_tsc;
1040 cur_tsc = rte_rdtsc();
1041 if (unlikely(cur_tsc - prev_tsc > MBUF_TABLE_DRAIN_TSC)) {
1044 RTE_LOG(DEBUG, VHOST_DATA,
1045 "TX queue drained after timeout with burst size %u\n",
1047 do_drain_mbuf_table(tx_q);
1051 static inline void __attribute__((always_inline))
1052 drain_eth_rx(struct vhost_dev *vdev)
1054 uint16_t rx_count, enqueue_count;
1055 struct virtio_net *dev = vdev->dev;
1056 struct rte_mbuf *pkts[MAX_PKT_BURST];
1058 rx_count = rte_eth_rx_burst(ports[0], vdev->vmdq_rx_q,
1059 pkts, MAX_PKT_BURST);
1064 * When "enable_retry" is set, here we wait and retry when there
1065 * is no enough free slots in the queue to hold @rx_count packets,
1066 * to diminish packet loss.
1069 unlikely(rx_count > rte_vring_available_entries(dev,
1073 for (retry = 0; retry < burst_rx_retry_num; retry++) {
1074 rte_delay_us(burst_rx_delay_time);
1075 if (rx_count <= rte_vring_available_entries(dev,
1081 enqueue_count = rte_vhost_enqueue_burst(dev, VIRTIO_RXQ,
1084 uint64_t fh = dev->device_fh;
1086 rte_atomic64_add(&dev_statistics[fh].rx_total_atomic, rx_count);
1087 rte_atomic64_add(&dev_statistics[fh].rx_atomic, enqueue_count);
1090 free_pkts(pkts, rx_count);
1093 static inline void __attribute__((always_inline))
1094 drain_virtio_tx(struct vhost_dev *vdev)
1096 struct rte_mbuf *pkts[MAX_PKT_BURST];
1100 count = rte_vhost_dequeue_burst(vdev->dev, VIRTIO_TXQ, mbuf_pool,
1101 pkts, MAX_PKT_BURST);
1103 /* setup VMDq for the first packet */
1104 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && count) {
1105 if (vdev->remove || link_vmdq(vdev, pkts[0]) == -1)
1106 free_pkts(pkts, count);
1109 for (i = 0; i < count; ++i) {
1110 virtio_tx_route(vdev, pkts[i],
1111 vlan_tags[(uint16_t)vdev->dev->device_fh]);
1116 * Main function of vhost-switch. It basically does:
1118 * for each vhost device {
1121 * Which drains the host eth Rx queue linked to the vhost device,
1122 * and deliver all of them to guest virito Rx ring associated with
1123 * this vhost device.
1125 * - drain_virtio_tx()
1127 * Which drains the guest virtio Tx queue and deliver all of them
1128 * to the target, which could be another vhost device, or the
1129 * physical eth dev. The route is done in function "virtio_tx_route".
1133 switch_worker(void *arg __rte_unused)
1136 unsigned lcore_id = rte_lcore_id();
1137 struct vhost_dev *vdev;
1138 struct mbuf_table *tx_q;
1140 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
1142 tx_q = &lcore_tx_queue[lcore_id];
1143 for (i = 0; i < rte_lcore_count(); i++) {
1144 if (lcore_ids[i] == lcore_id) {
1151 drain_mbuf_table(tx_q);
1154 * Inform the configuration core that we have exited the
1155 * linked list and that no devices are in use if requested.
1157 if (lcore_info[lcore_id].dev_removal_flag == REQUEST_DEV_REMOVAL)
1158 lcore_info[lcore_id].dev_removal_flag = ACK_DEV_REMOVAL;
1161 * Process vhost devices
1163 TAILQ_FOREACH(vdev, &lcore_info[lcore_id].vdev_list, next) {
1164 if (unlikely(vdev->remove)) {
1166 vdev->ready = DEVICE_SAFE_REMOVE;
1170 if (likely(vdev->ready == DEVICE_RX))
1173 if (likely(!vdev->remove))
1174 drain_virtio_tx(vdev);
1182 * Remove a device from the specific data core linked list and from the
1183 * main linked list. Synchonization occurs through the use of the
1184 * lcore dev_removal_flag. Device is made volatile here to avoid re-ordering
1185 * of dev->remove=1 which can cause an infinite loop in the rte_pause loop.
1188 destroy_device (volatile struct virtio_net *dev)
1190 struct vhost_dev *vdev;
1193 dev->flags &= ~VIRTIO_DEV_RUNNING;
1195 vdev = (struct vhost_dev *)dev->priv;
1196 /*set the remove flag. */
1198 while(vdev->ready != DEVICE_SAFE_REMOVE) {
1202 TAILQ_REMOVE(&lcore_info[vdev->coreid].vdev_list, vdev, next);
1203 TAILQ_REMOVE(&vhost_dev_list, vdev, next);
1205 /* Set the dev_removal_flag on each lcore. */
1206 RTE_LCORE_FOREACH_SLAVE(lcore)
1207 lcore_info[lcore].dev_removal_flag = REQUEST_DEV_REMOVAL;
1210 * Once each core has set the dev_removal_flag to ACK_DEV_REMOVAL
1211 * we can be sure that they can no longer access the device removed
1212 * from the linked lists and that the devices are no longer in use.
1214 RTE_LCORE_FOREACH_SLAVE(lcore) {
1215 while (lcore_info[lcore].dev_removal_flag != ACK_DEV_REMOVAL)
1219 lcore_info[vdev->coreid].device_num--;
1221 RTE_LOG(INFO, VHOST_DATA,
1222 "(%" PRIu64 ") Device has been removed from data core\n",
1229 * A new device is added to a data core. First the device is added to the main linked list
1230 * and the allocated to a specific data core.
1233 new_device (struct virtio_net *dev)
1235 int lcore, core_add = 0;
1236 uint32_t device_num_min = num_devices;
1237 struct vhost_dev *vdev;
1239 vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
1241 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Couldn't allocate memory for vhost dev\n",
1248 TAILQ_INSERT_TAIL(&vhost_dev_list, vdev, next);
1250 = dev->device_fh * queues_per_pool + vmdq_queue_base;
1252 /*reset ready flag*/
1253 vdev->ready = DEVICE_MAC_LEARNING;
1256 /* Find a suitable lcore to add the device. */
1257 RTE_LCORE_FOREACH_SLAVE(lcore) {
1258 if (lcore_info[lcore].device_num < device_num_min) {
1259 device_num_min = lcore_info[lcore].device_num;
1263 vdev->coreid = core_add;
1265 TAILQ_INSERT_TAIL(&lcore_info[vdev->coreid].vdev_list, vdev, next);
1266 lcore_info[vdev->coreid].device_num++;
1268 /* Initialize device stats */
1269 memset(&dev_statistics[dev->device_fh], 0, sizeof(struct device_statistics));
1271 /* Disable notifications. */
1272 rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
1273 rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
1274 dev->flags |= VIRTIO_DEV_RUNNING;
1276 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n", dev->device_fh, vdev->coreid);
1282 * These callback allow devices to be added to the data core when configuration
1283 * has been fully complete.
1285 static const struct virtio_net_device_ops virtio_net_device_ops =
1287 .new_device = new_device,
1288 .destroy_device = destroy_device,
1292 * This is a thread will wake up after a period to print stats if the user has
1298 struct vhost_dev *vdev;
1299 uint64_t tx_dropped, rx_dropped;
1300 uint64_t tx, tx_total, rx, rx_total;
1302 const char clr[] = { 27, '[', '2', 'J', '\0' };
1303 const char top_left[] = { 27, '[', '1', ';', '1', 'H','\0' };
1306 sleep(enable_stats);
1308 /* Clear screen and move to top left */
1309 printf("%s%s", clr, top_left);
1311 printf("\nDevice statistics ====================================");
1313 TAILQ_FOREACH(vdev, &vhost_dev_list, next) {
1314 device_fh = vdev->dev->device_fh;
1315 tx_total = dev_statistics[device_fh].tx_total;
1316 tx = dev_statistics[device_fh].tx;
1317 tx_dropped = tx_total - tx;
1318 rx_total = rte_atomic64_read(
1319 &dev_statistics[device_fh].rx_total_atomic);
1320 rx = rte_atomic64_read(
1321 &dev_statistics[device_fh].rx_atomic);
1322 rx_dropped = rx_total - rx;
1324 printf("\nStatistics for device %"PRIu32" ------------------------------"
1325 "\nTX total: %"PRIu64""
1326 "\nTX dropped: %"PRIu64""
1327 "\nTX successful: %"PRIu64""
1328 "\nRX total: %"PRIu64""
1329 "\nRX dropped: %"PRIu64""
1330 "\nRX successful: %"PRIu64"",
1339 printf("\n======================================================\n");
1343 /* When we receive a INT signal, unregister vhost driver */
1345 sigint_handler(__rte_unused int signum)
1347 /* Unregister vhost driver. */
1348 int ret = rte_vhost_driver_unregister((char *)&dev_basename);
1350 rte_exit(EXIT_FAILURE, "vhost driver unregister failure.\n");
1355 * While creating an mbuf pool, one key thing is to figure out how
1356 * many mbuf entries is enough for our use. FYI, here are some
1359 * - Each rx queue would reserve @nr_rx_desc mbufs at queue setup stage
1361 * - For each switch core (A CPU core does the packet switch), we need
1362 * also make some reservation for receiving the packets from virtio
1363 * Tx queue. How many is enough depends on the usage. It's normally
1364 * a simple calculation like following:
1366 * MAX_PKT_BURST * max packet size / mbuf size
1368 * So, we definitely need allocate more mbufs when TSO is enabled.
1370 * - Similarly, for each switching core, we should serve @nr_rx_desc
1371 * mbufs for receiving the packets from physical NIC device.
1373 * - We also need make sure, for each switch core, we have allocated
1374 * enough mbufs to fill up the mbuf cache.
1377 create_mbuf_pool(uint16_t nr_port, uint32_t nr_switch_core, uint32_t mbuf_size,
1378 uint32_t nr_queues, uint32_t nr_rx_desc, uint32_t nr_mbuf_cache)
1381 uint32_t nr_mbufs_per_core;
1382 uint32_t mtu = 1500;
1389 nr_mbufs_per_core = (mtu + mbuf_size) * MAX_PKT_BURST /
1390 (mbuf_size - RTE_PKTMBUF_HEADROOM) * MAX_PKT_BURST;
1391 nr_mbufs_per_core += nr_rx_desc;
1392 nr_mbufs_per_core = RTE_MAX(nr_mbufs_per_core, nr_mbuf_cache);
1394 nr_mbufs = nr_queues * nr_rx_desc;
1395 nr_mbufs += nr_mbufs_per_core * nr_switch_core;
1396 nr_mbufs *= nr_port;
1398 mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL", nr_mbufs,
1399 nr_mbuf_cache, 0, mbuf_size,
1401 if (mbuf_pool == NULL)
1402 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1406 * Main function, does initialisation and calls the per-lcore functions. The CUSE
1407 * device is also registered here to handle the IOCTLs.
1410 main(int argc, char *argv[])
1412 unsigned lcore_id, core_id = 0;
1413 unsigned nb_ports, valid_num_ports;
1416 static pthread_t tid;
1417 char thread_name[RTE_MAX_THREAD_NAME_LEN];
1419 signal(SIGINT, sigint_handler);
1422 ret = rte_eal_init(argc, argv);
1424 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1428 /* parse app arguments */
1429 ret = us_vhost_parse_args(argc, argv);
1431 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1433 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id ++)
1434 TAILQ_INIT(&lcore_info[lcore_id].vdev_list);
1436 if (rte_lcore_is_enabled(lcore_id))
1437 lcore_ids[core_id ++] = lcore_id;
1439 if (rte_lcore_count() > RTE_MAX_LCORE)
1440 rte_exit(EXIT_FAILURE,"Not enough cores\n");
1442 /* Get the number of physical ports. */
1443 nb_ports = rte_eth_dev_count();
1444 if (nb_ports > RTE_MAX_ETHPORTS)
1445 nb_ports = RTE_MAX_ETHPORTS;
1448 * Update the global var NUM_PORTS and global array PORTS
1449 * and get value of var VALID_NUM_PORTS according to system ports number
1451 valid_num_ports = check_ports_num(nb_ports);
1453 if ((valid_num_ports == 0) || (valid_num_ports > MAX_SUP_PORTS)) {
1454 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
1455 "but only %u port can be enabled\n",num_ports, MAX_SUP_PORTS);
1460 * FIXME: here we are trying to allocate mbufs big enough for
1461 * @MAX_QUEUES, but the truth is we're never going to use that
1462 * many queues here. We probably should only do allocation for
1463 * those queues we are going to use.
1465 create_mbuf_pool(valid_num_ports, rte_lcore_count() - 1, MBUF_DATA_SIZE,
1466 MAX_QUEUES, RTE_TEST_RX_DESC_DEFAULT, MBUF_CACHE_SIZE);
1468 if (vm2vm_mode == VM2VM_HARDWARE) {
1469 /* Enable VT loop back to let L2 switch to do it. */
1470 vmdq_conf_default.rx_adv_conf.vmdq_rx_conf.enable_loop_back = 1;
1471 RTE_LOG(DEBUG, VHOST_CONFIG,
1472 "Enable loop back for L2 switch in vmdq.\n");
1475 /* initialize all ports */
1476 for (portid = 0; portid < nb_ports; portid++) {
1477 /* skip ports that are not enabled */
1478 if ((enabled_port_mask & (1 << portid)) == 0) {
1479 RTE_LOG(INFO, VHOST_PORT,
1480 "Skipping disabled port %d\n", portid);
1483 if (port_init(portid) != 0)
1484 rte_exit(EXIT_FAILURE,
1485 "Cannot initialize network ports\n");
1488 /* Initialize device stats */
1489 memset(&dev_statistics, 0, sizeof(dev_statistics));
1491 /* Enable stats if the user option is set. */
1493 ret = pthread_create(&tid, NULL, (void *)print_stats, NULL);
1495 rte_exit(EXIT_FAILURE,
1496 "Cannot create print-stats thread\n");
1498 /* Set thread_name for aid in debugging. */
1499 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN, "print-stats");
1500 ret = rte_thread_setname(tid, thread_name);
1502 RTE_LOG(ERR, VHOST_CONFIG,
1503 "Cannot set print-stats name\n");
1506 /* Launch all data cores. */
1507 RTE_LCORE_FOREACH_SLAVE(lcore_id)
1508 rte_eal_remote_launch(switch_worker, NULL, lcore_id);
1511 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1513 /* Register vhost(cuse or user) driver to handle vhost messages. */
1514 ret = rte_vhost_driver_register((char *)&dev_basename);
1516 rte_exit(EXIT_FAILURE, "vhost driver register failure.\n");
1518 rte_vhost_driver_callback_register(&virtio_net_device_ops);
1520 /* Start CUSE session. */
1521 rte_vhost_driver_session_start();