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>
51 #include <rte_malloc.h>
52 #include <rte_virtio_net.h>
56 #include "vxlan_setup.h"
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 (nb_switching_cores * MAX_PKT_BURST) +\
66 (nb_switching_cores * \
67 RTE_TEST_TX_DESC_DEFAULT) +\
68 (nb_switching_cores * MBUF_CACHE_SIZE))
70 #define MBUF_CACHE_SIZE 128
71 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
73 #define MAX_PKT_BURST 32 /* Max burst size for RX/TX */
74 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
76 /* Defines how long we wait between retries on RX */
77 #define BURST_RX_WAIT_US 15
79 #define BURST_RX_RETRIES 4 /* Number of retries on RX. */
81 #define JUMBO_FRAME_MAX_SIZE 0x2600
83 /* State of virtio device. */
84 #define DEVICE_MAC_LEARNING 0
86 #define DEVICE_SAFE_REMOVE 2
88 /* Config_core_flag status definitions. */
89 #define REQUEST_DEV_REMOVAL 1
90 #define ACK_DEV_REMOVAL 0
92 /* Configurable number of RX/TX ring descriptors */
93 #define RTE_TEST_RX_DESC_DEFAULT 1024
94 #define RTE_TEST_TX_DESC_DEFAULT 512
96 /* Get first 4 bytes in mbuf headroom. */
97 #define MBUF_HEADROOM_UINT32(mbuf) (*(uint32_t *)((uint8_t *)(mbuf) \
98 + sizeof(struct rte_mbuf)))
100 #define INVALID_PORT_ID 0xFF
102 /* Size of buffers used for snprintfs. */
103 #define MAX_PRINT_BUFF 6072
105 /* Maximum character device basename size. */
106 #define MAX_BASENAME_SZ 20
108 /* Maximum long option length for option parsing. */
109 #define MAX_LONG_OPT_SZ 64
111 /* Used to compare MAC addresses. */
112 #define MAC_ADDR_CMP 0xFFFFFFFFFFFFULL
114 #define CMD_LINE_OPT_NB_DEVICES "nb-devices"
115 #define CMD_LINE_OPT_RX_RETRY "rx-retry"
116 #define CMD_LINE_OPT_RX_RETRY_DELAY "rx-retry-delay"
117 #define CMD_LINE_OPT_RX_RETRY_NUM "rx-retry-num"
118 #define CMD_LINE_OPT_STATS "stats"
119 #define CMD_LINE_OPT_DEV_BASENAME "dev-basename"
121 /* mask of enabled ports */
122 static uint32_t enabled_port_mask;
124 /*Number of switching cores enabled*/
125 static uint32_t nb_switching_cores;
127 /* number of devices/queues to support*/
128 uint16_t nb_devices = 2;
130 /* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
131 #define MAX_RING_DESC 4096
134 struct rte_mempool *pool;
135 struct rte_ring *ring;
137 } vpool_array[MAX_QUEUES+MAX_QUEUES];
139 /* overlay packet operation */
140 struct ol_switch_ops overlay_options = {
141 .port_configure = vxlan_port_init,
142 .tunnel_setup = vxlan_link,
143 .tunnel_destroy = vxlan_unlink,
144 .tx_handle = vxlan_tx_pkts,
145 .rx_handle = vxlan_rx_pkts,
146 .param_handle = NULL,
150 uint32_t enable_stats = 0;
151 /* Enable retries on RX. */
152 static uint32_t enable_retry = 1;
153 /* Specify timeout (in useconds) between retries on RX. */
154 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
155 /* Specify the number of retries on RX. */
156 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
158 /* Character device basename. Can be set by user. */
159 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
161 static unsigned lcore_ids[RTE_MAX_LCORE];
162 uint8_t ports[RTE_MAX_ETHPORTS];
164 static unsigned nb_ports; /**< The number of ports specified in command line */
166 /* ethernet addresses of ports */
167 struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
169 /* heads for the main used and free linked lists for the data path. */
170 static struct virtio_net_data_ll *ll_root_used;
171 static struct virtio_net_data_ll *ll_root_free;
174 * Array of data core structures containing information on
175 * individual core linked lists.
177 static struct lcore_info lcore_info[RTE_MAX_LCORE];
179 /* Used for queueing bursts of TX packets. */
183 struct rte_mbuf *m_table[MAX_PKT_BURST];
186 /* TX queue for each data core. */
187 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
189 struct device_statistics dev_statistics[MAX_DEVICES];
192 * Set character device basename.
195 us_vhost_parse_basename(const char *q_arg)
197 /* parse number string */
198 if (strlen(q_arg) >= MAX_BASENAME_SZ)
201 snprintf((char *)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
207 * Parse the portmask provided at run time.
210 parse_portmask(const char *portmask)
215 /* parse hexadecimal string */
216 pm = strtoul(portmask, &end, 16);
217 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
227 * Parse num options at run time.
230 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
235 /* parse unsigned int string */
236 num = strtoul(q_arg, &end, 10);
237 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
240 if (num > max_valid_value)
250 tep_termination_usage(const char *prgname)
252 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
253 " --nb-devices[1-64]: The number of virtIO device\n"
254 " -p PORTMASK: Set mask for ports to be used by application\n"
255 " --rx-retry [0|1]: disable/enable(default) retries on rx."
256 " Enable retry if destintation queue is full\n"
257 " --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX."
258 " This makes effect only if retries on rx enabled\n"
259 " --rx-retry-num [0-N]: the number of retries on rx."
260 " This makes effect only if retries on rx enabled\n"
261 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
262 " --dev-basename: The basename to be used for the character device.\n",
267 * Parse the arguments given in the command line of the application.
270 tep_termination_parse_args(int argc, char **argv)
275 const char *prgname = argv[0];
276 static struct option long_option[] = {
277 {CMD_LINE_OPT_NB_DEVICES, required_argument, NULL, 0},
278 {CMD_LINE_OPT_RX_RETRY, required_argument, NULL, 0},
279 {CMD_LINE_OPT_RX_RETRY_DELAY, required_argument, NULL, 0},
280 {CMD_LINE_OPT_RX_RETRY_NUM, required_argument, NULL, 0},
281 {CMD_LINE_OPT_STATS, required_argument, NULL, 0},
282 {CMD_LINE_OPT_DEV_BASENAME, required_argument, NULL, 0},
286 /* Parse command line */
287 while ((opt = getopt_long(argc, argv, "p:",
288 long_option, &option_index)) != EOF) {
292 enabled_port_mask = parse_portmask(optarg);
293 if (enabled_port_mask == 0) {
294 RTE_LOG(INFO, VHOST_CONFIG,
295 "Invalid portmask\n");
296 tep_termination_usage(prgname);
301 if (!strncmp(long_option[option_index].name,
302 CMD_LINE_OPT_NB_DEVICES,
303 sizeof(CMD_LINE_OPT_NB_DEVICES))) {
304 ret = parse_num_opt(optarg, MAX_DEVICES);
306 RTE_LOG(INFO, VHOST_CONFIG,
307 "Invalid argument for nb-devices [0-%d]\n",
309 tep_termination_usage(prgname);
315 /* Enable/disable retries on RX. */
316 if (!strncmp(long_option[option_index].name,
317 CMD_LINE_OPT_RX_RETRY,
318 sizeof(CMD_LINE_OPT_RX_RETRY))) {
319 ret = parse_num_opt(optarg, 1);
321 RTE_LOG(INFO, VHOST_CONFIG,
322 "Invalid argument for rx-retry [0|1]\n");
323 tep_termination_usage(prgname);
329 /* Specify the retries delay time (in useconds) on RX.*/
330 if (!strncmp(long_option[option_index].name,
331 CMD_LINE_OPT_RX_RETRY_DELAY,
332 sizeof(CMD_LINE_OPT_RX_RETRY_DELAY))) {
333 ret = parse_num_opt(optarg, INT32_MAX);
335 RTE_LOG(INFO, VHOST_CONFIG,
336 "Invalid argument for rx-retry-delay [0-N]\n");
337 tep_termination_usage(prgname);
340 burst_rx_delay_time = ret;
343 /* Specify the retries number on RX. */
344 if (!strncmp(long_option[option_index].name,
345 CMD_LINE_OPT_RX_RETRY_NUM,
346 sizeof(CMD_LINE_OPT_RX_RETRY_NUM))) {
347 ret = parse_num_opt(optarg, INT32_MAX);
349 RTE_LOG(INFO, VHOST_CONFIG,
350 "Invalid argument for rx-retry-num [0-N]\n");
351 tep_termination_usage(prgname);
354 burst_rx_retry_num = ret;
357 /* Enable/disable stats. */
358 if (!strncmp(long_option[option_index].name,
360 sizeof(CMD_LINE_OPT_STATS))) {
361 ret = parse_num_opt(optarg, INT32_MAX);
363 RTE_LOG(INFO, VHOST_CONFIG,
364 "Invalid argument for stats [0..N]\n");
365 tep_termination_usage(prgname);
371 /* Set character device basename. */
372 if (!strncmp(long_option[option_index].name,
373 CMD_LINE_OPT_DEV_BASENAME,
374 sizeof(CMD_LINE_OPT_DEV_BASENAME))) {
375 if (us_vhost_parse_basename(optarg) == -1) {
376 RTE_LOG(INFO, VHOST_CONFIG,
377 "Invalid argument for character "
378 "device basename (Max %d characters)\n",
380 tep_termination_usage(prgname);
387 /* Invalid option - print options. */
389 tep_termination_usage(prgname);
394 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
395 if (enabled_port_mask & (1 << i))
396 ports[nb_ports++] = (uint8_t)i;
399 if ((nb_ports == 0) || (nb_ports > MAX_SUP_PORTS)) {
400 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
401 "but only %u port can be enabled\n", nb_ports,
410 * Update the global var NB_PORTS and array PORTS
411 * according to system ports number and return valid ports number
414 check_ports_num(unsigned max_nb_ports)
416 unsigned valid_nb_ports = nb_ports;
419 if (nb_ports > max_nb_ports) {
420 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) "
421 " exceeds total system port number(%u)\n",
422 nb_ports, max_nb_ports);
423 nb_ports = max_nb_ports;
426 for (portid = 0; portid < nb_ports; portid++) {
427 if (ports[portid] >= max_nb_ports) {
428 RTE_LOG(INFO, VHOST_PORT,
429 "\nSpecified port ID(%u) exceeds max "
430 " system port ID(%u)\n",
431 ports[portid], (max_nb_ports - 1));
432 ports[portid] = INVALID_PORT_ID;
436 return valid_nb_ports;
440 * This function routes the TX packet to the correct interface. This may be a local device
441 * or the physical port.
443 static inline void __attribute__((always_inline))
444 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
446 struct mbuf_table *tx_q;
447 struct rte_mbuf **m_table;
448 unsigned len, ret = 0;
449 const uint16_t lcore_id = rte_lcore_id();
450 struct virtio_net *dev = vdev->dev;
452 LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is external\n",
455 /* Add packet to the port tx queue */
456 tx_q = &lcore_tx_queue[lcore_id];
459 tx_q->m_table[len] = m;
462 dev_statistics[dev->device_fh].tx_total++;
463 dev_statistics[dev->device_fh].tx++;
466 if (unlikely(len == MAX_PKT_BURST)) {
467 m_table = (struct rte_mbuf **)tx_q->m_table;
468 ret = overlay_options.tx_handle(ports[0],
469 (uint16_t)tx_q->txq_id, m_table,
470 (uint16_t)tx_q->len);
472 /* Free any buffers not handled by TX and update
475 if (unlikely(ret < len)) {
477 rte_pktmbuf_free(m_table[ret]);
478 } while (++ret < len);
489 * This function is called by each data core. It handles all
490 * RX/TX registered with the core. For TX the specific lcore
491 * linked list is used. For RX, MAC addresses are compared
492 * with all devices in the main linked list.
495 switch_worker(__rte_unused void *arg)
497 struct rte_mempool *mbuf_pool = arg;
498 struct virtio_net *dev = NULL;
499 struct vhost_dev *vdev = NULL;
500 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
501 struct virtio_net_data_ll *dev_ll;
502 struct mbuf_table *tx_q;
503 volatile struct lcore_ll_info *lcore_ll;
504 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
505 / US_PER_S * BURST_TX_DRAIN_US;
506 uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
508 const uint16_t lcore_id = rte_lcore_id();
509 const uint16_t num_cores = (uint16_t)rte_lcore_count();
510 uint16_t rx_count = 0;
514 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
515 lcore_ll = lcore_info[lcore_id].lcore_ll;
518 tx_q = &lcore_tx_queue[lcore_id];
519 for (i = 0; i < num_cores; i++) {
520 if (lcore_ids[i] == lcore_id) {
527 cur_tsc = rte_rdtsc();
529 * TX burst queue drain
531 diff_tsc = cur_tsc - prev_tsc;
532 if (unlikely(diff_tsc > drain_tsc)) {
535 LOG_DEBUG(VHOST_DATA, "TX queue drained after "
536 "timeout with burst size %u\n",
538 ret = overlay_options.tx_handle(ports[0],
539 (uint16_t)tx_q->txq_id,
540 (struct rte_mbuf **)tx_q->m_table,
541 (uint16_t)tx_q->len);
542 if (unlikely(ret < tx_q->len)) {
544 rte_pktmbuf_free(tx_q->m_table[ret]);
545 } while (++ret < tx_q->len);
555 rte_prefetch0(lcore_ll->ll_root_used);
558 * Inform the configuration core that we have exited
559 * the linked list and that no devices are
560 * in use if requested.
562 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
563 lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
568 dev_ll = lcore_ll->ll_root_used;
570 while (dev_ll != NULL) {
574 if (unlikely(vdev->remove)) {
575 dev_ll = dev_ll->next;
576 overlay_options.tunnel_destroy(vdev);
577 vdev->ready = DEVICE_SAFE_REMOVE;
580 if (likely(vdev->ready == DEVICE_RX)) {
581 /* Handle guest RX */
582 rx_count = rte_eth_rx_burst(ports[0],
583 vdev->rx_q, pkts_burst, MAX_PKT_BURST);
587 * Retry is enabled and the queue is
588 * full then we wait and retry to
589 * avoid packet loss. Here MAX_PKT_BURST
590 * must be less than virtio queue size
592 if (enable_retry && unlikely(rx_count >
593 rte_vring_available_entries(dev, VIRTIO_RXQ))) {
594 for (retry = 0; retry < burst_rx_retry_num;
596 rte_delay_us(burst_rx_delay_time);
597 if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
602 ret_count = overlay_options.rx_handle(dev, pkts_burst, rx_count);
605 &dev_statistics[dev->device_fh].rx_total_atomic,
608 &dev_statistics[dev->device_fh].rx_atomic, ret_count);
610 while (likely(rx_count)) {
612 rte_pktmbuf_free(pkts_burst[rx_count]);
618 if (likely(!vdev->remove)) {
620 tx_count = rte_vhost_dequeue_burst(dev,
621 VIRTIO_TXQ, mbuf_pool,
622 pkts_burst, MAX_PKT_BURST);
623 /* If this is the first received packet we need to learn the MAC */
624 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
626 (overlay_options.tunnel_setup(vdev, pkts_burst[0]) == -1)) {
628 rte_pktmbuf_free(pkts_burst[--tx_count]);
632 virtio_tx_route(vdev, pkts_burst[--tx_count]);
635 /* move to the next device in the list */
636 dev_ll = dev_ll->next;
644 * Add an entry to a used linked list. A free entry must first be found
645 * in the free linked list using get_data_ll_free_entry();
648 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
649 struct virtio_net_data_ll *ll_dev)
651 struct virtio_net_data_ll *ll = *ll_root_addr;
653 /* Set next as NULL and use a compiler barrier to avoid reordering. */
655 rte_compiler_barrier();
657 /* If ll == NULL then this is the first device. */
659 /* Increment to the tail of the linked list. */
660 while (ll->next != NULL)
665 *ll_root_addr = ll_dev;
670 * Remove an entry from a used linked list. The entry must then be added to
671 * the free linked list using put_data_ll_free_entry().
674 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
675 struct virtio_net_data_ll *ll_dev,
676 struct virtio_net_data_ll *ll_dev_last)
678 struct virtio_net_data_ll *ll = *ll_root_addr;
680 if (unlikely((ll == NULL) || (ll_dev == NULL)))
684 *ll_root_addr = ll_dev->next;
686 if (likely(ll_dev_last != NULL))
687 ll_dev_last->next = ll_dev->next;
689 RTE_LOG(ERR, VHOST_CONFIG,
690 "Remove entry form ll failed.\n");
694 * Find and return an entry from the free linked list.
696 static struct virtio_net_data_ll *
697 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
699 struct virtio_net_data_ll *ll_free = *ll_root_addr;
700 struct virtio_net_data_ll *ll_dev;
706 *ll_root_addr = ll_free->next;
712 * Place an entry back on to the free linked list.
715 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
716 struct virtio_net_data_ll *ll_dev)
718 struct virtio_net_data_ll *ll_free = *ll_root_addr;
723 ll_dev->next = ll_free;
724 *ll_root_addr = ll_dev;
728 * Creates a linked list of a given size.
730 static struct virtio_net_data_ll *
731 alloc_data_ll(uint32_t size)
733 struct virtio_net_data_ll *ll_new;
736 /* Malloc and then chain the linked list. */
737 ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
738 if (ll_new == NULL) {
739 RTE_LOG(ERR, VHOST_CONFIG,
740 "Failed to allocate memory for ll_new.\n");
744 for (i = 0; i < size - 1; i++) {
745 ll_new[i].vdev = NULL;
746 ll_new[i].next = &ll_new[i+1];
748 ll_new[i].next = NULL;
754 * Create the main linked list along with each individual cores
755 * linked list. A used and a free list are created to manage entries.
762 RTE_LCORE_FOREACH_SLAVE(lcore) {
763 lcore_info[lcore].lcore_ll =
764 malloc(sizeof(struct lcore_ll_info));
765 if (lcore_info[lcore].lcore_ll == NULL) {
766 RTE_LOG(ERR, VHOST_CONFIG,
767 "Failed to allocate memory for lcore_ll.\n");
771 lcore_info[lcore].lcore_ll->device_num = 0;
772 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
773 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
774 if (nb_devices % nb_switching_cores)
775 lcore_info[lcore].lcore_ll->ll_root_free =
776 alloc_data_ll((nb_devices / nb_switching_cores)
779 lcore_info[lcore].lcore_ll->ll_root_free =
780 alloc_data_ll(nb_devices / nb_switching_cores);
783 /* Allocate devices up to a maximum of MAX_DEVICES. */
784 ll_root_free = alloc_data_ll(MIN((nb_devices), MAX_DEVICES));
790 * Remove a device from the specific data core linked list and
791 * from the main linked list. Synchonization occurs through the use
792 * of the lcore dev_removal_flag. Device is made volatile here
793 * to avoid re-ordering of dev->remove=1 which can cause an infinite
794 * loop in the rte_pause loop.
797 destroy_device(volatile struct virtio_net *dev)
799 struct virtio_net_data_ll *ll_lcore_dev_cur;
800 struct virtio_net_data_ll *ll_main_dev_cur;
801 struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
802 struct virtio_net_data_ll *ll_main_dev_last = NULL;
803 struct vhost_dev *vdev;
806 dev->flags &= ~VIRTIO_DEV_RUNNING;
808 vdev = (struct vhost_dev *)dev->priv;
810 /* set the remove flag. */
812 while (vdev->ready != DEVICE_SAFE_REMOVE)
815 /* Search for entry to be removed from lcore ll */
816 ll_lcore_dev_cur = lcore_info[vdev->coreid].lcore_ll->ll_root_used;
817 while (ll_lcore_dev_cur != NULL) {
818 if (ll_lcore_dev_cur->vdev == vdev) {
821 ll_lcore_dev_last = ll_lcore_dev_cur;
822 ll_lcore_dev_cur = ll_lcore_dev_cur->next;
826 if (ll_lcore_dev_cur == NULL) {
827 RTE_LOG(ERR, VHOST_CONFIG,
828 "(%"PRIu64") Failed to find the dev to be destroy.\n",
833 /* Search for entry to be removed from main ll */
834 ll_main_dev_cur = ll_root_used;
835 ll_main_dev_last = NULL;
836 while (ll_main_dev_cur != NULL) {
837 if (ll_main_dev_cur->vdev == vdev) {
840 ll_main_dev_last = ll_main_dev_cur;
841 ll_main_dev_cur = ll_main_dev_cur->next;
845 /* Remove entries from the lcore and main ll. */
846 rm_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
847 ll_lcore_dev_cur, ll_lcore_dev_last);
848 rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
850 /* Set the dev_removal_flag on each lcore. */
851 RTE_LCORE_FOREACH_SLAVE(lcore) {
852 lcore_info[lcore].lcore_ll->dev_removal_flag =
857 * Once each core has set the dev_removal_flag to
858 * ACK_DEV_REMOVAL we can be sure that they can no longer access
859 * the device removed from the linked lists and that the devices
860 * are no longer in use.
862 RTE_LCORE_FOREACH_SLAVE(lcore) {
863 while (lcore_info[lcore].lcore_ll->dev_removal_flag
868 /* Add the entries back to the lcore and main free ll.*/
869 put_data_ll_free_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_free,
871 put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
873 /* Decrement number of device on the lcore. */
874 lcore_info[vdev->coreid].lcore_ll->device_num--;
876 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been removed "
877 "from data core\n", dev->device_fh);
884 * A new device is added to a data core. First the device is added
885 * to the main linked list and the allocated to a specific data core.
888 new_device(struct virtio_net *dev)
890 struct virtio_net_data_ll *ll_dev;
891 int lcore, core_add = 0;
892 uint32_t device_num_min = nb_devices;
893 struct vhost_dev *vdev;
895 vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
897 RTE_LOG(INFO, VHOST_DATA,
898 "(%"PRIu64") Couldn't allocate memory for vhost dev\n",
904 /* Add device to main ll */
905 ll_dev = get_data_ll_free_entry(&ll_root_free);
906 if (ll_dev == NULL) {
907 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in"
908 " linked list Device limit of %d devices per core"
909 " has been reached\n", dev->device_fh, nb_devices);
910 if (vdev->regions_hpa)
911 rte_free(vdev->regions_hpa);
916 add_data_ll_entry(&ll_root_used, ll_dev);
917 vdev->rx_q = dev->device_fh;
919 /* reset ready flag */
920 vdev->ready = DEVICE_MAC_LEARNING;
923 /* Find a suitable lcore to add the device. */
924 RTE_LCORE_FOREACH_SLAVE(lcore) {
925 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
926 device_num_min = lcore_info[lcore].lcore_ll->device_num;
930 /* Add device to lcore ll */
931 ll_dev = get_data_ll_free_entry(&lcore_info[core_add].lcore_ll->ll_root_free);
932 if (ll_dev == NULL) {
933 RTE_LOG(INFO, VHOST_DATA,
934 "(%"PRIu64") Failed to add device to data core\n",
936 vdev->ready = DEVICE_SAFE_REMOVE;
938 rte_free(vdev->regions_hpa);
943 vdev->coreid = core_add;
945 add_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
948 /* Initialize device stats */
949 memset(&dev_statistics[dev->device_fh], 0,
950 sizeof(struct device_statistics));
952 /* Disable notifications. */
953 rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
954 rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
955 lcore_info[vdev->coreid].lcore_ll->device_num++;
956 dev->flags |= VIRTIO_DEV_RUNNING;
958 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n",
959 dev->device_fh, vdev->coreid);
965 * These callback allow devices to be added to the data core when configuration
966 * has been fully complete.
968 static const struct virtio_net_device_ops virtio_net_device_ops = {
969 .new_device = new_device,
970 .destroy_device = destroy_device,
974 * This is a thread will wake up after a period to print stats if the user has
980 struct virtio_net_data_ll *dev_ll;
981 uint64_t tx_dropped, rx_dropped;
982 uint64_t tx, tx_total, rx, rx_total;
984 const char clr[] = { 27, '[', '2', 'J', '\0' };
985 const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
990 /* Clear screen and move to top left */
991 printf("%s%s", clr, top_left);
993 printf("\nDevice statistics ================================");
995 dev_ll = ll_root_used;
996 while (dev_ll != NULL) {
997 device_fh = (uint32_t)dev_ll->vdev->dev->device_fh;
998 tx_total = dev_statistics[device_fh].tx_total;
999 tx = dev_statistics[device_fh].tx;
1000 tx_dropped = tx_total - tx;
1002 rx_total = rte_atomic64_read(
1003 &dev_statistics[device_fh].rx_total_atomic);
1004 rx = rte_atomic64_read(
1005 &dev_statistics[device_fh].rx_atomic);
1006 rx_dropped = rx_total - rx;
1008 printf("\nStatistics for device %"PRIu32" ----------"
1009 "\nTX total: %"PRIu64""
1010 "\nTX dropped: %"PRIu64""
1011 "\nTX successful: %"PRIu64""
1012 "\nRX total: %"PRIu64""
1013 "\nRX dropped: %"PRIu64""
1014 "\nRX successful: %"PRIu64"",
1023 dev_ll = dev_ll->next;
1025 printf("\n================================================\n");
1030 * Main function, does initialisation and calls the per-lcore functions. The CUSE
1031 * device is also registered here to handle the IOCTLs.
1034 main(int argc, char *argv[])
1036 struct rte_mempool *mbuf_pool = NULL;
1037 unsigned lcore_id, core_id = 0;
1038 unsigned nb_ports, valid_nb_ports;
1042 static pthread_t tid;
1045 ret = rte_eal_init(argc, argv);
1047 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1051 /* parse app arguments */
1052 ret = tep_termination_parse_args(argc, argv);
1054 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1056 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
1057 if (rte_lcore_is_enabled(lcore_id))
1058 lcore_ids[core_id++] = lcore_id;
1060 /* set the number of swithcing cores available */
1061 nb_switching_cores = rte_lcore_count()-1;
1063 /* Get the number of physical ports. */
1064 nb_ports = rte_eth_dev_count();
1065 if (nb_ports > RTE_MAX_ETHPORTS)
1066 nb_ports = RTE_MAX_ETHPORTS;
1069 * Update the global var NB_PORTS and global array PORTS
1070 * and get value of var VALID_NB_PORTS according to system ports number
1072 valid_nb_ports = check_ports_num(nb_ports);
1074 if ((valid_nb_ports == 0) || (valid_nb_ports > MAX_SUP_PORTS)) {
1075 rte_exit(EXIT_FAILURE, "Current enabled port number is %u,"
1076 "but only %u port can be enabled\n", nb_ports,
1079 /* Create the mbuf pool. */
1080 mbuf_pool = rte_mempool_create(
1084 MBUF_SIZE, MBUF_CACHE_SIZE,
1085 sizeof(struct rte_pktmbuf_pool_private),
1086 rte_pktmbuf_pool_init, NULL,
1087 rte_pktmbuf_init, NULL,
1088 rte_socket_id(), 0);
1089 if (mbuf_pool == NULL)
1090 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1092 for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
1093 vpool_array[queue_id].pool = mbuf_pool;
1095 /* Set log level. */
1096 rte_set_log_level(LOG_LEVEL);
1098 /* initialize all ports */
1099 for (portid = 0; portid < nb_ports; portid++) {
1100 /* skip ports that are not enabled */
1101 if ((enabled_port_mask & (1 << portid)) == 0) {
1102 RTE_LOG(INFO, VHOST_PORT,
1103 "Skipping disabled port %d\n", portid);
1106 if (overlay_options.port_configure(portid, mbuf_pool) != 0)
1107 rte_exit(EXIT_FAILURE,
1108 "Cannot initialize network ports\n");
1111 /* Initialise all linked lists. */
1112 if (init_data_ll() == -1)
1113 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1115 /* Initialize device stats */
1116 memset(&dev_statistics, 0, sizeof(dev_statistics));
1118 /* Enable stats if the user option is set. */
1120 pthread_create(&tid, NULL, (void *)print_stats, NULL);
1122 /* Launch all data cores. */
1123 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1124 rte_eal_remote_launch(switch_worker,
1125 mbuf_pool, lcore_id);
1127 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1129 /* Register CUSE device to handle IOCTLs. */
1130 ret = rte_vhost_driver_register((char *)&dev_basename);
1132 rte_exit(EXIT_FAILURE, "CUSE device setup failure.\n");
1134 rte_vhost_driver_callback_register(&virtio_net_device_ops);
1136 /* Start CUSE session. */
1137 rte_vhost_driver_session_start();