4 * Copyright(c) 2010-2015 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>
53 #include "virtio-net.h"
54 #include "xen_vhost.h"
56 #define MAX_QUEUES 128
58 /* the maximum number of external ports supported */
59 #define MAX_SUP_PORTS 1
62 * Calculate the number of buffers needed per port
64 #define NUM_MBUFS_PER_PORT ((MAX_QUEUES*RTE_TEST_RX_DESC_DEFAULT) + \
65 (num_switching_cores*MAX_PKT_BURST) + \
66 (num_switching_cores*RTE_TEST_TX_DESC_DEFAULT) +\
67 (num_switching_cores*MBUF_CACHE_SIZE))
69 #define MBUF_CACHE_SIZE 64
72 * RX and TX Prefetch, Host, and Write-back threshold values should be
73 * carefully set for optimal performance. Consult the network
74 * controller's datasheet and supporting DPDK documentation for guidance
75 * on how these parameters should be set.
77 #define RX_PTHRESH 8 /* Default values of RX prefetch threshold reg. */
78 #define RX_HTHRESH 8 /* Default values of RX host threshold reg. */
79 #define RX_WTHRESH 4 /* Default values of RX write-back threshold reg. */
82 * These default values are optimized for use with the Intel(R) 82599 10 GbE
83 * Controller and the DPDK ixgbe PMD. Consider using other values for other
84 * network controllers and/or network drivers.
86 #define TX_PTHRESH 36 /* Default values of TX prefetch threshold reg. */
87 #define TX_HTHRESH 0 /* Default values of TX host threshold reg. */
88 #define TX_WTHRESH 0 /* Default values of TX write-back threshold reg. */
90 #define MAX_PKT_BURST 32 /* Max burst size for RX/TX */
91 #define MAX_MRG_PKT_BURST 16 /* Max burst for merge buffers. Set to 1 due to performance issue. */
92 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
94 /* State of virtio device. */
95 #define DEVICE_NOT_READY 0
96 #define DEVICE_READY 1
97 #define DEVICE_SAFE_REMOVE 2
99 /* Config_core_flag status definitions. */
100 #define REQUEST_DEV_REMOVAL 1
101 #define ACK_DEV_REMOVAL 0
103 /* Configurable number of RX/TX ring descriptors */
104 #define RTE_TEST_RX_DESC_DEFAULT 128
105 #define RTE_TEST_TX_DESC_DEFAULT 512
107 #define INVALID_PORT_ID 0xFF
109 /* Max number of devices. Limited by vmdq. */
110 #define MAX_DEVICES 64
112 /* Size of buffers used for snprintfs. */
113 #define MAX_PRINT_BUFF 6072
116 /* Maximum long option length for option parsing. */
117 #define MAX_LONG_OPT_SZ 64
119 /* Used to compare MAC addresses. */
120 #define MAC_ADDR_CMP 0xFFFFFFFFFFFF
122 /* mask of enabled ports */
123 static uint32_t enabled_port_mask = 0;
125 /*Number of switching cores enabled*/
126 static uint32_t num_switching_cores = 0;
128 /* number of devices/queues to support*/
129 static uint32_t num_queues = 0;
130 uint32_t num_devices = 0;
132 /* Enable VM2VM communications. If this is disabled then the MAC address compare is skipped. */
133 static uint32_t enable_vm2vm = 1;
135 static uint32_t enable_stats = 0;
137 /* empty vmdq configuration structure. Filled in programatically */
138 static const struct rte_eth_conf vmdq_conf_default = {
140 .mq_mode = ETH_MQ_RX_VMDQ_ONLY,
142 .header_split = 0, /**< Header Split disabled */
143 .hw_ip_checksum = 0, /**< IP checksum offload disabled */
144 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
146 * It is necessary for 1G NIC such as I350,
147 * this fixes bug of ipv4 forwarding in guest can't
148 * forward pakets from one virtio dev to another virtio dev.
150 .hw_vlan_strip = 1, /**< VLAN strip enabled. */
151 .jumbo_frame = 0, /**< Jumbo Frame Support disabled */
152 .hw_strip_crc = 0, /**< CRC stripped by hardware */
156 .mq_mode = ETH_MQ_TX_NONE,
160 * should be overridden separately in code with
164 .nb_queue_pools = ETH_8_POOLS,
165 .enable_default_pool = 0,
168 .pool_map = {{0, 0},},
173 static unsigned lcore_ids[RTE_MAX_LCORE];
174 static uint8_t ports[RTE_MAX_ETHPORTS];
175 static unsigned num_ports = 0; /**< The number of ports specified in command line */
177 const uint16_t vlan_tags[] = {
178 1000, 1001, 1002, 1003, 1004, 1005, 1006, 1007,
179 1008, 1009, 1010, 1011, 1012, 1013, 1014, 1015,
180 1016, 1017, 1018, 1019, 1020, 1021, 1022, 1023,
181 1024, 1025, 1026, 1027, 1028, 1029, 1030, 1031,
182 1032, 1033, 1034, 1035, 1036, 1037, 1038, 1039,
183 1040, 1041, 1042, 1043, 1044, 1045, 1046, 1047,
184 1048, 1049, 1050, 1051, 1052, 1053, 1054, 1055,
185 1056, 1057, 1058, 1059, 1060, 1061, 1062, 1063,
188 /* ethernet addresses of ports */
189 static struct ether_addr vmdq_ports_eth_addr[RTE_MAX_ETHPORTS];
191 /* heads for the main used and free linked lists for the data path. */
192 static struct virtio_net_data_ll *ll_root_used = NULL;
193 static struct virtio_net_data_ll *ll_root_free = NULL;
195 /* Array of data core structures containing information on individual core linked lists. */
196 static struct lcore_info lcore_info[RTE_MAX_LCORE];
198 /* Used for queueing bursts of TX packets. */
202 struct rte_mbuf *m_table[MAX_PKT_BURST];
205 /* TX queue for each data core. */
206 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
208 /* Vlan header struct used to insert vlan tags on TX. */
210 unsigned char h_dest[ETH_ALEN];
211 unsigned char h_source[ETH_ALEN];
214 __be16 h_vlan_encapsulated_proto;
217 /* Header lengths. */
219 #define VLAN_ETH_HLEN 18
221 /* Per-device statistics struct */
222 struct device_statistics {
224 rte_atomic64_t rx_total;
227 } __rte_cache_aligned;
228 struct device_statistics dev_statistics[MAX_DEVICES];
231 * Builds up the correct configuration for VMDQ VLAN pool map
232 * according to the pool & queue limits.
235 get_eth_conf(struct rte_eth_conf *eth_conf, uint32_t num_devices)
237 struct rte_eth_vmdq_rx_conf 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;
244 for (i = 0; i < conf.nb_pool_maps; i++) {
245 conf.pool_map[i].vlan_id = vlan_tags[ i ];
246 conf.pool_map[i].pools = (1UL << i);
249 (void)(rte_memcpy(eth_conf, &vmdq_conf_default, sizeof(*eth_conf)));
250 (void)(rte_memcpy(ð_conf->rx_adv_conf.vmdq_rx_conf, &conf,
251 sizeof(eth_conf->rx_adv_conf.vmdq_rx_conf)));
256 * Validate the device number according to the max pool number gotten form dev_info
257 * If the device number is invalid, give the error message and return -1.
258 * Each device must have its own pool.
261 validate_num_devices(uint32_t max_nb_devices)
263 if (num_devices > max_nb_devices) {
264 RTE_LOG(ERR, VHOST_PORT, "invalid number of devices\n");
271 * Initialises a given port using global settings and with the rx buffers
272 * coming from the mbuf_pool passed as parameter
275 port_init(uint8_t port, struct rte_mempool *mbuf_pool)
277 struct rte_eth_dev_info dev_info;
278 struct rte_eth_rxconf *rxconf;
279 struct rte_eth_conf port_conf;
280 uint16_t rx_rings, tx_rings = (uint16_t)rte_lcore_count();
281 const uint16_t rx_ring_size = RTE_TEST_RX_DESC_DEFAULT, tx_ring_size = RTE_TEST_TX_DESC_DEFAULT;
285 /* The max pool number from dev_info will be used to validate the pool number specified in cmd line */
286 rte_eth_dev_info_get (port, &dev_info);
288 /*configure the number of supported virtio devices based on VMDQ limits */
289 num_devices = dev_info.max_vmdq_pools;
290 num_queues = dev_info.max_rx_queues;
292 retval = validate_num_devices(MAX_DEVICES);
296 /* Get port configuration. */
297 retval = get_eth_conf(&port_conf, num_devices);
301 if (port >= rte_eth_dev_count()) return -1;
303 rx_rings = (uint16_t)num_queues,
304 /* Configure ethernet device. */
305 retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
309 rte_eth_dev_info_get(port, &dev_info);
310 rxconf = &dev_info.default_rxconf;
311 rxconf->rx_drop_en = 1;
312 /* Setup the queues. */
313 for (q = 0; q < rx_rings; q ++) {
314 retval = rte_eth_rx_queue_setup(port, q, rx_ring_size,
315 rte_eth_dev_socket_id(port), rxconf,
320 for (q = 0; q < tx_rings; q ++) {
321 retval = rte_eth_tx_queue_setup(port, q, tx_ring_size,
322 rte_eth_dev_socket_id(port),
328 /* Start the device. */
329 retval = rte_eth_dev_start(port);
333 rte_eth_macaddr_get(port, &vmdq_ports_eth_addr[port]);
334 RTE_LOG(INFO, VHOST_PORT, "Max virtio devices supported: %u\n", num_devices);
335 RTE_LOG(INFO, VHOST_PORT, "Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
336 " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
338 vmdq_ports_eth_addr[port].addr_bytes[0],
339 vmdq_ports_eth_addr[port].addr_bytes[1],
340 vmdq_ports_eth_addr[port].addr_bytes[2],
341 vmdq_ports_eth_addr[port].addr_bytes[3],
342 vmdq_ports_eth_addr[port].addr_bytes[4],
343 vmdq_ports_eth_addr[port].addr_bytes[5]);
349 * Parse the portmask provided at run time.
352 parse_portmask(const char *portmask)
359 /* parse hexadecimal string */
360 pm = strtoul(portmask, &end, 16);
361 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
372 * Parse num options at run time.
375 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
382 /* parse unsigned int string */
383 num = strtoul(q_arg, &end, 10);
384 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
387 if (num > max_valid_value)
398 us_vhost_usage(const char *prgname)
400 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK --vm2vm [0|1] --stats [0-N] --nb-devices ND\n"
401 " -p PORTMASK: Set mask for ports to be used by application\n"
402 " --vm2vm [0|1]: disable/enable(default) vm2vm comms\n"
403 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n",
408 * Parse the arguments given in the command line of the application.
411 us_vhost_parse_args(int argc, char **argv)
416 const char *prgname = argv[0];
417 static struct option long_option[] = {
418 {"vm2vm", required_argument, NULL, 0},
419 {"stats", required_argument, NULL, 0},
423 /* Parse command line */
424 while ((opt = getopt_long(argc, argv, "p:",long_option, &option_index)) != EOF) {
428 enabled_port_mask = parse_portmask(optarg);
429 if (enabled_port_mask == 0) {
430 RTE_LOG(INFO, VHOST_CONFIG, "Invalid portmask\n");
431 us_vhost_usage(prgname);
437 /* Enable/disable vm2vm comms. */
438 if (!strncmp(long_option[option_index].name, "vm2vm", MAX_LONG_OPT_SZ)) {
439 ret = parse_num_opt(optarg, 1);
441 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for vm2vm [0|1]\n");
442 us_vhost_usage(prgname);
449 /* Enable/disable stats. */
450 if (!strncmp(long_option[option_index].name, "stats", MAX_LONG_OPT_SZ)) {
451 ret = parse_num_opt(optarg, INT32_MAX);
453 RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for stats [0..N]\n");
454 us_vhost_usage(prgname);
462 /* Invalid option - print options. */
464 us_vhost_usage(prgname);
469 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
470 if (enabled_port_mask & (1 << i))
471 ports[num_ports++] = (uint8_t)i;
474 if ((num_ports == 0) || (num_ports > MAX_SUP_PORTS)) {
475 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
476 "but only %u port can be enabled\n",num_ports, MAX_SUP_PORTS);
484 * Update the global var NUM_PORTS and array PORTS according to system ports number
485 * and return valid ports number
487 static unsigned check_ports_num(unsigned nb_ports)
489 unsigned valid_num_ports = num_ports;
492 if (num_ports > nb_ports) {
493 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) exceeds total system port number(%u)\n",
494 num_ports, nb_ports);
495 num_ports = nb_ports;
498 for (portid = 0; portid < num_ports; portid ++) {
499 if (ports[portid] >= nb_ports) {
500 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port ID(%u) exceeds max system port ID(%u)\n",
501 ports[portid], (nb_ports - 1));
502 ports[portid] = INVALID_PORT_ID;
506 return valid_num_ports;
510 * Function to convert guest physical addresses to vhost virtual addresses. This
511 * is used to convert virtio buffer addresses.
513 static inline uint64_t __attribute__((always_inline))
514 gpa_to_vva(struct virtio_net *dev, uint64_t guest_pa)
516 struct virtio_memory_regions *region;
518 uint64_t vhost_va = 0;
520 for (regionidx = 0; regionidx < dev->mem->nregions; regionidx++) {
521 region = &dev->mem->regions[regionidx];
522 if ((guest_pa >= region->guest_phys_address) &&
523 (guest_pa <= region->guest_phys_address_end)) {
524 vhost_va = region->address_offset + guest_pa;
528 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") GPA %p| VVA %p\n",
529 dev->device_fh, (void*)(uintptr_t)guest_pa, (void*)(uintptr_t)vhost_va);
535 * This function adds buffers to the virtio devices RX virtqueue. Buffers can
536 * be received from the physical port or from another virtio device. A packet
537 * count is returned to indicate the number of packets that were succesfully
538 * added to the RX queue.
540 static inline uint32_t __attribute__((always_inline))
541 virtio_dev_rx(struct virtio_net *dev, struct rte_mbuf **pkts, uint32_t count)
543 struct vhost_virtqueue *vq;
544 struct vring_desc *desc;
545 struct rte_mbuf *buff;
546 /* The virtio_hdr is initialised to 0. */
547 struct virtio_net_hdr_mrg_rxbuf virtio_hdr = {{0,0,0,0,0,0},0};
548 uint64_t buff_addr = 0;
549 uint64_t buff_hdr_addr = 0;
550 uint32_t head[MAX_PKT_BURST], packet_len = 0;
551 uint32_t head_idx, packet_success = 0;
552 uint16_t avail_idx, res_cur_idx;
553 uint16_t res_base_idx, res_end_idx;
554 uint16_t free_entries;
558 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") virtio_dev_rx()\n", dev->device_fh);
559 vq = dev->virtqueue_rx;
560 count = (count > MAX_PKT_BURST) ? MAX_PKT_BURST : count;
561 /* As many data cores may want access to available buffers, they need to be reserved. */
564 res_base_idx = vq->last_used_idx_res;
566 avail_idx = *((volatile uint16_t *)&vq->avail->idx);
568 free_entries = (avail_idx - res_base_idx);
570 /*check that we have enough buffers*/
571 if (unlikely(count > free_entries))
572 count = free_entries;
577 res_end_idx = res_base_idx + count;
578 /* vq->last_used_idx_res is atomically updated. */
579 success = rte_atomic16_cmpset(&vq->last_used_idx_res, res_base_idx,
581 } while (unlikely(success == 0));
582 res_cur_idx = res_base_idx;
583 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") Current Index %d| End Index %d\n",
584 dev->device_fh, res_cur_idx, res_end_idx);
586 /* Prefetch available ring to retrieve indexes. */
587 rte_prefetch0(&vq->avail->ring[res_cur_idx & (vq->size - 1)]);
589 /* Retrieve all of the head indexes first to avoid caching issues. */
590 for (head_idx = 0; head_idx < count; head_idx++)
591 head[head_idx] = vq->avail->ring[(res_cur_idx + head_idx) & (vq->size - 1)];
593 /*Prefetch descriptor index. */
594 rte_prefetch0(&vq->desc[head[packet_success]]);
596 while (res_cur_idx != res_end_idx) {
597 /* Get descriptor from available ring */
598 desc = &vq->desc[head[packet_success]];
599 /* Prefetch descriptor address. */
602 buff = pkts[packet_success];
604 /* Convert from gpa to vva (guest physical addr -> vhost virtual addr) */
605 buff_addr = gpa_to_vva(dev, desc->addr);
606 /* Prefetch buffer address. */
607 rte_prefetch0((void*)(uintptr_t)buff_addr);
610 /* Copy virtio_hdr to packet and increment buffer address */
611 buff_hdr_addr = buff_addr;
612 packet_len = rte_pktmbuf_data_len(buff) + vq->vhost_hlen;
615 * If the descriptors are chained the header and data are placed in
618 if (desc->flags & VRING_DESC_F_NEXT) {
619 desc->len = vq->vhost_hlen;
620 desc = &vq->desc[desc->next];
621 /* Buffer address translation. */
622 buff_addr = gpa_to_vva(dev, desc->addr);
623 desc->len = rte_pktmbuf_data_len(buff);
625 buff_addr += vq->vhost_hlen;
626 desc->len = packet_len;
630 /* Update used ring with desc information */
631 vq->used->ring[res_cur_idx & (vq->size - 1)].id = head[packet_success];
632 vq->used->ring[res_cur_idx & (vq->size - 1)].len = packet_len;
634 /* Copy mbuf data to buffer */
635 userdata = rte_pktmbuf_mtod(buff, void *);
636 rte_memcpy((void *)(uintptr_t)buff_addr, userdata, rte_pktmbuf_data_len(buff));
641 /* mergeable is disabled then a header is required per buffer. */
642 rte_memcpy((void *)(uintptr_t)buff_hdr_addr, (const void *)&virtio_hdr, vq->vhost_hlen);
643 if (res_cur_idx < res_end_idx) {
644 /* Prefetch descriptor index. */
645 rte_prefetch0(&vq->desc[head[packet_success]]);
649 rte_compiler_barrier();
651 /* Wait until it's our turn to add our buffer to the used ring. */
652 while (unlikely(vq->last_used_idx != res_base_idx))
655 *(volatile uint16_t *)&vq->used->idx += count;
657 vq->last_used_idx = res_end_idx;
663 * Compares a packet destination MAC address to a device MAC address.
665 static inline int __attribute__((always_inline))
666 ether_addr_cmp(struct ether_addr *ea, struct ether_addr *eb)
668 return ((*(uint64_t *)ea ^ *(uint64_t *)eb) & MAC_ADDR_CMP) == 0;
672 * This function registers mac along with a
673 * vlan tag to a VMDQ.
676 link_vmdq(struct virtio_net *dev)
679 struct virtio_net_data_ll *dev_ll;
681 dev_ll = ll_root_used;
683 while (dev_ll != NULL) {
684 if ((dev != dev_ll->dev) && ether_addr_cmp(&dev->mac_address, &dev_ll->dev->mac_address)) {
685 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") WARNING: This device is using an existing MAC address and has not been registered.\n", dev->device_fh);
688 dev_ll = dev_ll->next;
691 /* vlan_tag currently uses the device_id. */
692 dev->vlan_tag = vlan_tags[dev->device_fh];
693 dev->vmdq_rx_q = dev->device_fh * (num_queues/num_devices);
695 /* Print out VMDQ registration info. */
696 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") MAC_ADDRESS %02x:%02x:%02x:%02x:%02x:%02x and VLAN_TAG %d registered\n",
698 dev->mac_address.addr_bytes[0], dev->mac_address.addr_bytes[1],
699 dev->mac_address.addr_bytes[2], dev->mac_address.addr_bytes[3],
700 dev->mac_address.addr_bytes[4], dev->mac_address.addr_bytes[5],
703 /* Register the MAC address. */
704 ret = rte_eth_dev_mac_addr_add(ports[0], &dev->mac_address, (uint32_t)dev->device_fh);
706 RTE_LOG(ERR, VHOST_DATA, "(%"PRIu64") Failed to add device MAC address to VMDQ\n",
711 /* Enable stripping of the vlan tag as we handle routing. */
712 rte_eth_dev_set_vlan_strip_on_queue(ports[0], dev->vmdq_rx_q, 1);
714 rte_compiler_barrier();
715 /* Set device as ready for RX. */
716 dev->ready = DEVICE_READY;
722 * Removes MAC address and vlan tag from VMDQ. Ensures that nothing is adding buffers to the RX
723 * queue before disabling RX on the device.
726 unlink_vmdq(struct virtio_net *dev)
730 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
732 if (dev->ready == DEVICE_READY) {
733 /*clear MAC and VLAN settings*/
734 rte_eth_dev_mac_addr_remove(ports[0], &dev->mac_address);
735 for (i = 0; i < 6; i++)
736 dev->mac_address.addr_bytes[i] = 0;
740 /*Clear out the receive buffers*/
741 rx_count = rte_eth_rx_burst(ports[0],
742 (uint16_t)dev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
745 for (i = 0; i < rx_count; i++)
746 rte_pktmbuf_free(pkts_burst[i]);
748 rx_count = rte_eth_rx_burst(ports[0],
749 (uint16_t)dev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
752 dev->ready = DEVICE_NOT_READY;
757 * Check if the packet destination MAC address is for a local device. If so then put
758 * the packet on that devices RX queue. If not then return.
760 static inline unsigned __attribute__((always_inline))
761 virtio_tx_local(struct virtio_net *dev, struct rte_mbuf *m)
763 struct virtio_net_data_ll *dev_ll;
764 struct ether_hdr *pkt_hdr;
767 pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
769 /*get the used devices list*/
770 dev_ll = ll_root_used;
772 while (dev_ll != NULL) {
773 if (likely(dev_ll->dev->ready == DEVICE_READY) && ether_addr_cmp(&(pkt_hdr->d_addr),
774 &dev_ll->dev->mac_address)) {
776 /* Drop the packet if the TX packet is destined for the TX device. */
777 if (dev_ll->dev->device_fh == dev->device_fh) {
778 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") TX: "
779 "Source and destination MAC addresses are the same. "
780 "Dropping packet.\n",
781 dev_ll->dev->device_fh);
786 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") TX: "
787 "MAC address is local\n", dev_ll->dev->device_fh);
789 if (dev_ll->dev->remove) {
790 /*drop the packet if the device is marked for removal*/
791 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") "
792 "Device is marked for removal\n",
793 dev_ll->dev->device_fh);
795 /*send the packet to the local virtio device*/
796 ret = virtio_dev_rx(dev_ll->dev, &m, 1);
798 rte_atomic64_add(&dev_statistics[dev_ll->dev->device_fh].rx_total, 1);
799 rte_atomic64_add(&dev_statistics[dev_ll->dev->device_fh].rx, ret);
800 dev_statistics[dev->device_fh].tx_total++;
801 dev_statistics[dev->device_fh].tx += ret;
807 dev_ll = dev_ll->next;
814 * This function routes the TX packet to the correct interface. This may be a local device
815 * or the physical port.
817 static inline void __attribute__((always_inline))
818 virtio_tx_route(struct virtio_net* dev, struct rte_mbuf *m, struct rte_mempool *mbuf_pool, uint16_t vlan_tag)
820 struct mbuf_table *tx_q;
821 struct vlan_ethhdr *vlan_hdr;
822 struct rte_mbuf **m_table;
823 struct rte_mbuf *mbuf;
825 const uint16_t lcore_id = rte_lcore_id();
827 /*check if destination is local VM*/
828 if (enable_vm2vm && (virtio_tx_local(dev, m) == 0)) {
832 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") TX: "
833 "MAC address is external\n", dev->device_fh);
835 /*Add packet to the port tx queue*/
836 tx_q = &lcore_tx_queue[lcore_id];
839 /* Allocate an mbuf and populate the structure. */
840 mbuf = rte_pktmbuf_alloc(mbuf_pool);
844 mbuf->data_len = m->data_len + VLAN_HLEN;
845 mbuf->pkt_len = mbuf->data_len;
847 /* Copy ethernet header to mbuf. */
848 rte_memcpy(rte_pktmbuf_mtod(mbuf, void*),
849 rte_pktmbuf_mtod(m, const void*), ETH_HLEN);
852 /* Setup vlan header. Bytes need to be re-ordered for network with htons()*/
853 vlan_hdr = rte_pktmbuf_mtod(mbuf, struct vlan_ethhdr *);
854 vlan_hdr->h_vlan_encapsulated_proto = vlan_hdr->h_vlan_proto;
855 vlan_hdr->h_vlan_proto = htons(ETH_P_8021Q);
856 vlan_hdr->h_vlan_TCI = htons(vlan_tag);
858 /* Copy the remaining packet contents to the mbuf. */
859 rte_memcpy(rte_pktmbuf_mtod_offset(mbuf, void *, VLAN_ETH_HLEN),
860 rte_pktmbuf_mtod_offset(m, const void *, ETH_HLEN),
861 (m->data_len - ETH_HLEN));
862 tx_q->m_table[len] = mbuf;
865 dev_statistics[dev->device_fh].tx_total++;
866 dev_statistics[dev->device_fh].tx++;
869 if (unlikely(len == MAX_PKT_BURST)) {
870 m_table = (struct rte_mbuf **)tx_q->m_table;
871 ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id, m_table, (uint16_t) len);
872 /* Free any buffers not handled by TX and update the port stats. */
873 if (unlikely(ret < len)) {
875 rte_pktmbuf_free(m_table[ret]);
876 } while (++ret < len);
886 static inline void __attribute__((always_inline))
887 virtio_dev_tx(struct virtio_net* dev, struct rte_mempool *mbuf_pool)
890 struct vhost_virtqueue *vq;
891 struct vring_desc *desc;
892 uint64_t buff_addr = 0;
893 uint32_t head[MAX_PKT_BURST];
896 uint16_t free_entries, packet_success = 0;
899 vq = dev->virtqueue_tx;
900 avail_idx = *((volatile uint16_t *)&vq->avail->idx);
902 /* If there are no available buffers then return. */
903 if (vq->last_used_idx == avail_idx)
906 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") virtio_dev_tx()\n",
909 /* Prefetch available ring to retrieve head indexes. */
910 rte_prefetch0(&vq->avail->ring[vq->last_used_idx & (vq->size - 1)]);
912 /*get the number of free entries in the ring*/
913 free_entries = avail_idx - vq->last_used_idx;
914 free_entries = unlikely(free_entries < MAX_PKT_BURST) ? free_entries : MAX_PKT_BURST;
916 RTE_LOG_DP(DEBUG, VHOST_DATA, "(%" PRIu64 ") Buffers available %d\n",
917 dev->device_fh, free_entries);
918 /* Retrieve all of the head indexes first to avoid caching issues. */
919 for (i = 0; i < free_entries; i++)
920 head[i] = vq->avail->ring[(vq->last_used_idx + i) & (vq->size - 1)];
922 /* Prefetch descriptor index. */
923 rte_prefetch0(&vq->desc[head[packet_success]]);
925 while (packet_success < free_entries) {
926 desc = &vq->desc[head[packet_success]];
927 /* Prefetch descriptor address. */
930 if (packet_success < (free_entries - 1)) {
931 /* Prefetch descriptor index. */
932 rte_prefetch0(&vq->desc[head[packet_success+1]]);
935 /* Update used index buffer information. */
936 used_idx = vq->last_used_idx & (vq->size - 1);
937 vq->used->ring[used_idx].id = head[packet_success];
938 vq->used->ring[used_idx].len = 0;
940 /* Discard first buffer as it is the virtio header */
941 desc = &vq->desc[desc->next];
943 /* Buffer address translation. */
944 buff_addr = gpa_to_vva(dev, desc->addr);
945 /* Prefetch buffer address. */
946 rte_prefetch0((void*)(uintptr_t)buff_addr);
948 /* Setup dummy mbuf. This is copied to a real mbuf if transmitted out the physical port. */
949 m.data_len = desc->len;
953 virtio_tx_route(dev, &m, mbuf_pool, 0);
959 rte_compiler_barrier();
960 vq->used->idx += packet_success;
961 /* Kick guest if required. */
965 * This function is called by each data core. It handles all RX/TX registered with the
966 * core. For TX the specific lcore linked list is used. For RX, MAC addresses are compared
967 * with all devices in the main linked list.
970 switch_worker(__attribute__((unused)) void *arg)
972 struct rte_mempool *mbuf_pool = arg;
973 struct virtio_net *dev = NULL;
974 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
975 struct virtio_net_data_ll *dev_ll;
976 struct mbuf_table *tx_q;
977 volatile struct lcore_ll_info *lcore_ll;
978 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
979 uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
981 const uint16_t lcore_id = rte_lcore_id();
982 const uint16_t num_cores = (uint16_t)rte_lcore_count();
983 uint16_t rx_count = 0;
985 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started \n", lcore_id);
986 lcore_ll = lcore_info[lcore_id].lcore_ll;
989 tx_q = &lcore_tx_queue[lcore_id];
990 for (i = 0; i < num_cores; i ++) {
991 if (lcore_ids[i] == lcore_id) {
998 cur_tsc = rte_rdtsc();
1000 * TX burst queue drain
1002 diff_tsc = cur_tsc - prev_tsc;
1003 if (unlikely(diff_tsc > drain_tsc)) {
1006 RTE_LOG_DP(DEBUG, VHOST_DATA,
1007 "TX queue drained after timeout with burst size %u\n",
1010 /*Tx any packets in the queue*/
1011 ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id,
1012 (struct rte_mbuf **)tx_q->m_table,
1013 (uint16_t)tx_q->len);
1014 if (unlikely(ret < tx_q->len)) {
1016 rte_pktmbuf_free(tx_q->m_table[ret]);
1017 } while (++ret < tx_q->len);
1028 * Inform the configuration core that we have exited the linked list and that no devices are
1029 * in use if requested.
1031 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
1032 lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
1037 dev_ll = lcore_ll->ll_root_used;
1039 while (dev_ll != NULL) {
1040 /*get virtio device ID*/
1043 if (unlikely(dev->remove)) {
1044 dev_ll = dev_ll->next;
1046 dev->ready = DEVICE_SAFE_REMOVE;
1049 if (likely(dev->ready == DEVICE_READY)) {
1051 rx_count = rte_eth_rx_burst(ports[0],
1052 (uint16_t)dev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
1055 ret_count = virtio_dev_rx(dev, pkts_burst, rx_count);
1057 rte_atomic64_add(&dev_statistics[dev_ll->dev->device_fh].rx_total, rx_count);
1058 rte_atomic64_add(&dev_statistics[dev_ll->dev->device_fh].rx, ret_count);
1060 while (likely(rx_count)) {
1062 rte_pktmbuf_free_seg(pkts_burst[rx_count]);
1068 if (likely(!dev->remove))
1070 virtio_dev_tx(dev, mbuf_pool);
1072 /*move to the next device in the list*/
1073 dev_ll = dev_ll->next;
1081 * Add an entry to a used linked list. A free entry must first be found in the free linked list
1082 * using get_data_ll_free_entry();
1085 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr, struct virtio_net_data_ll *ll_dev)
1087 struct virtio_net_data_ll *ll = *ll_root_addr;
1089 /* Set next as NULL and use a compiler barrier to avoid reordering. */
1090 ll_dev->next = NULL;
1091 rte_compiler_barrier();
1093 /* If ll == NULL then this is the first device. */
1095 /* Increment to the tail of the linked list. */
1096 while ((ll->next != NULL) )
1101 *ll_root_addr = ll_dev;
1106 * Remove an entry from a used linked list. The entry must then be added to the free linked list
1107 * using put_data_ll_free_entry().
1110 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr, struct virtio_net_data_ll *ll_dev, struct virtio_net_data_ll *ll_dev_last)
1112 struct virtio_net_data_ll *ll = *ll_root_addr;
1115 *ll_root_addr = ll_dev->next;
1117 ll_dev_last->next = ll_dev->next;
1121 * Find and return an entry from the free linked list.
1123 static struct virtio_net_data_ll *
1124 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
1126 struct virtio_net_data_ll *ll_free = *ll_root_addr;
1127 struct virtio_net_data_ll *ll_dev;
1129 if (ll_free == NULL)
1133 *ll_root_addr = ll_free->next;
1139 * Place an entry back on to the free linked list.
1142 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr, struct virtio_net_data_ll *ll_dev)
1144 struct virtio_net_data_ll *ll_free = *ll_root_addr;
1146 ll_dev->next = ll_free;
1147 *ll_root_addr = ll_dev;
1151 * Creates a linked list of a given size.
1153 static struct virtio_net_data_ll *
1154 alloc_data_ll(uint32_t size)
1156 struct virtio_net_data_ll *ll_new;
1159 /* Malloc and then chain the linked list. */
1160 ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
1161 if (ll_new == NULL) {
1162 RTE_LOG(ERR, VHOST_CONFIG, "Failed to allocate memory for ll_new.\n");
1166 for (i = 0; i < size - 1; i++) {
1167 ll_new[i].dev = NULL;
1168 ll_new[i].next = &ll_new[i+1];
1170 ll_new[i].next = NULL;
1176 * Create the main linked list along with each individual cores linked list. A used and a free list
1177 * are created to manage entries.
1184 RTE_LCORE_FOREACH_SLAVE(lcore) {
1185 lcore_info[lcore].lcore_ll = malloc(sizeof(struct lcore_ll_info));
1186 if (lcore_info[lcore].lcore_ll == NULL) {
1187 RTE_LOG(ERR, VHOST_CONFIG, "Failed to allocate memory for lcore_ll.\n");
1191 lcore_info[lcore].lcore_ll->device_num = 0;
1192 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
1193 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
1194 if (num_devices % num_switching_cores)
1195 lcore_info[lcore].lcore_ll->ll_root_free = alloc_data_ll((num_devices / num_switching_cores) + 1);
1197 lcore_info[lcore].lcore_ll->ll_root_free = alloc_data_ll(num_devices / num_switching_cores);
1200 /* Allocate devices up to a maximum of MAX_DEVICES. */
1201 ll_root_free = alloc_data_ll(MIN((num_devices), MAX_DEVICES));
1206 * Remove a device from the specific data core linked list and from the main linked list. The
1207 * rx/tx thread must be set the flag to indicate that it is safe to remove the device.
1211 destroy_device (volatile struct virtio_net *dev)
1213 struct virtio_net_data_ll *ll_lcore_dev_cur;
1214 struct virtio_net_data_ll *ll_main_dev_cur;
1215 struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
1216 struct virtio_net_data_ll *ll_main_dev_last = NULL;
1219 dev->flags &= ~VIRTIO_DEV_RUNNING;
1221 /*set the remove flag. */
1224 while(dev->ready != DEVICE_SAFE_REMOVE) {
1228 /* Search for entry to be removed from lcore ll */
1229 ll_lcore_dev_cur = lcore_info[dev->coreid].lcore_ll->ll_root_used;
1230 while (ll_lcore_dev_cur != NULL) {
1231 if (ll_lcore_dev_cur->dev == dev) {
1234 ll_lcore_dev_last = ll_lcore_dev_cur;
1235 ll_lcore_dev_cur = ll_lcore_dev_cur->next;
1239 /* Search for entry to be removed from main ll */
1240 ll_main_dev_cur = ll_root_used;
1241 ll_main_dev_last = NULL;
1242 while (ll_main_dev_cur != NULL) {
1243 if (ll_main_dev_cur->dev == dev) {
1246 ll_main_dev_last = ll_main_dev_cur;
1247 ll_main_dev_cur = ll_main_dev_cur->next;
1251 if (ll_lcore_dev_cur == NULL || ll_main_dev_cur == NULL) {
1252 RTE_LOG(ERR, XENHOST, "%s: could find device in per_cpu list or main_list\n", __func__);
1256 /* Remove entries from the lcore and main ll. */
1257 rm_data_ll_entry(&lcore_info[ll_lcore_dev_cur->dev->coreid].lcore_ll->ll_root_used, ll_lcore_dev_cur, ll_lcore_dev_last);
1258 rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
1260 /* Set the dev_removal_flag on each lcore. */
1261 RTE_LCORE_FOREACH_SLAVE(lcore) {
1262 lcore_info[lcore].lcore_ll->dev_removal_flag = REQUEST_DEV_REMOVAL;
1266 * Once each core has set the dev_removal_flag to ACK_DEV_REMOVAL we can be sure that
1267 * they can no longer access the device removed from the linked lists and that the devices
1268 * are no longer in use.
1270 RTE_LCORE_FOREACH_SLAVE(lcore) {
1271 while (lcore_info[lcore].lcore_ll->dev_removal_flag != ACK_DEV_REMOVAL) {
1276 /* Add the entries back to the lcore and main free ll.*/
1277 put_data_ll_free_entry(&lcore_info[ll_lcore_dev_cur->dev->coreid].lcore_ll->ll_root_free, ll_lcore_dev_cur);
1278 put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
1280 /* Decrement number of device on the lcore. */
1281 lcore_info[ll_lcore_dev_cur->dev->coreid].lcore_ll->device_num--;
1283 RTE_LOG(INFO, VHOST_DATA, " #####(%"PRIu64") Device has been removed from data core\n", dev->device_fh);
1287 * A new device is added to a data core. First the device is added to the main linked list
1288 * and the allocated to a specific data core.
1291 new_device (struct virtio_net *dev)
1293 struct virtio_net_data_ll *ll_dev;
1294 int lcore, core_add = 0;
1295 uint32_t device_num_min = num_devices;
1297 /* Add device to main ll */
1298 ll_dev = get_data_ll_free_entry(&ll_root_free);
1299 if (ll_dev == NULL) {
1300 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in linked list. Device limit "
1301 "of %d devices per core has been reached\n",
1302 dev->device_fh, num_devices);
1306 add_data_ll_entry(&ll_root_used, ll_dev);
1308 /*reset ready flag*/
1309 dev->ready = DEVICE_NOT_READY;
1312 /* Find a suitable lcore to add the device. */
1313 RTE_LCORE_FOREACH_SLAVE(lcore) {
1314 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
1315 device_num_min = lcore_info[lcore].lcore_ll->device_num;
1319 /* Add device to lcore ll */
1320 ll_dev->dev->coreid = core_add;
1321 ll_dev = get_data_ll_free_entry(&lcore_info[ll_dev->dev->coreid].lcore_ll->ll_root_free);
1322 if (ll_dev == NULL) {
1323 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Failed to add device to data core\n", dev->device_fh);
1324 destroy_device(dev);
1328 add_data_ll_entry(&lcore_info[ll_dev->dev->coreid].lcore_ll->ll_root_used, ll_dev);
1330 /* Initialize device stats */
1331 memset(&dev_statistics[dev->device_fh], 0, sizeof(struct device_statistics));
1333 lcore_info[ll_dev->dev->coreid].lcore_ll->device_num++;
1334 dev->flags |= VIRTIO_DEV_RUNNING;
1336 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n", dev->device_fh, dev->coreid);
1344 * These callback allow devices to be added to the data core when configuration
1345 * has been fully complete.
1347 static const struct virtio_net_device_ops virtio_net_device_ops =
1349 .new_device = new_device,
1350 .destroy_device = destroy_device,
1354 * This is a thread will wake up after a period to print stats if the user has
1360 struct virtio_net_data_ll *dev_ll;
1361 uint64_t tx_dropped, rx_dropped;
1362 uint64_t tx, tx_total, rx, rx_total;
1364 const char clr[] = { 27, '[', '2', 'J', '\0' };
1365 const char top_left[] = { 27, '[', '1', ';', '1', 'H','\0' };
1368 sleep(enable_stats);
1370 /* Clear screen and move to top left */
1371 printf("%s%s", clr, top_left);
1373 printf("\nDevice statistics ====================================");
1375 dev_ll = ll_root_used;
1376 while (dev_ll != NULL) {
1377 device_fh = (uint32_t)dev_ll->dev->device_fh;
1378 tx_total = dev_statistics[device_fh].tx_total;
1379 tx = dev_statistics[device_fh].tx;
1380 tx_dropped = tx_total - tx;
1381 rx_total = rte_atomic64_read(&dev_statistics[device_fh].rx_total);
1382 rx = rte_atomic64_read(&dev_statistics[device_fh].rx);
1383 rx_dropped = rx_total - rx;
1385 printf("\nStatistics for device %"PRIu32" ------------------------------"
1386 "\nTX total: %"PRIu64""
1387 "\nTX dropped: %"PRIu64""
1388 "\nTX successful: %"PRIu64""
1389 "\nRX total: %"PRIu64""
1390 "\nRX dropped: %"PRIu64""
1391 "\nRX successful: %"PRIu64"",
1400 dev_ll = dev_ll->next;
1402 printf("\n======================================================\n");
1407 int init_virtio_net(struct virtio_net_device_ops const * const ops);
1410 * Main function, does initialisation and calls the per-lcore functions.
1413 main(int argc, char *argv[])
1415 struct rte_mempool *mbuf_pool;
1416 unsigned lcore_id, core_id = 0;
1417 unsigned nb_ports, valid_num_ports;
1420 static pthread_t tid;
1421 char thread_name[RTE_MAX_THREAD_NAME_LEN];
1424 ret = rte_eal_init(argc, argv);
1426 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1430 /* parse app arguments */
1431 ret = us_vhost_parse_args(argc, argv);
1433 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1435 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id ++)
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 /*set the number of swithcing cores available*/
1443 num_switching_cores = rte_lcore_count()-1;
1445 /* Get the number of physical ports. */
1446 nb_ports = rte_eth_dev_count();
1449 * Update the global var NUM_PORTS and global array PORTS
1450 * and get value of var VALID_NUM_PORTS according to system ports number
1452 valid_num_ports = check_ports_num(nb_ports);
1454 if ((valid_num_ports == 0) || (valid_num_ports > MAX_SUP_PORTS)) {
1455 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
1456 "but only %u port can be enabled\n",num_ports, MAX_SUP_PORTS);
1460 /* Create the mbuf pool. */
1461 mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL",
1462 NUM_MBUFS_PER_PORT * valid_num_ports, MBUF_CACHE_SIZE, 0,
1463 RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
1464 if (mbuf_pool == NULL)
1465 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1467 /* initialize all ports */
1468 for (portid = 0; portid < nb_ports; portid++) {
1469 /* skip ports that are not enabled */
1470 if ((enabled_port_mask & (1 << portid)) == 0) {
1471 RTE_LOG(INFO, VHOST_PORT, "Skipping disabled port %d\n", portid);
1474 if (port_init(portid, mbuf_pool) != 0)
1475 rte_exit(EXIT_FAILURE, "Cannot initialize network ports\n");
1478 /* Initialise all linked lists. */
1479 if (init_data_ll() == -1)
1480 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1482 /* Initialize device stats */
1483 memset(&dev_statistics, 0, sizeof(dev_statistics));
1485 /* Enable stats if the user option is set. */
1487 ret = pthread_create(&tid, NULL, (void *)print_stats, NULL);
1489 rte_exit(EXIT_FAILURE,
1490 "Cannot create print-stats thread\n");
1492 /* Set thread_name for aid in debugging. */
1493 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN, "print-xen-stats");
1494 ret = rte_thread_setname(tid, thread_name);
1496 RTE_LOG(DEBUG, VHOST_CONFIG,
1497 "Cannot set print-stats name\n");
1500 /* Launch all data cores. */
1501 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1502 rte_eal_remote_launch(switch_worker, mbuf_pool, lcore_id);
1505 init_virtio_xen(&virtio_net_device_ops);
1507 virtio_monitor_loop();