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*/
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];
140 uint32_t enable_stats = 0;
141 /* Enable retries on RX. */
142 static uint32_t enable_retry = 1;
143 /* Specify timeout (in useconds) between retries on RX. */
144 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
145 /* Specify the number of retries on RX. */
146 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
148 /* Character device basename. Can be set by user. */
149 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
151 static unsigned lcore_ids[RTE_MAX_LCORE];
152 uint8_t ports[RTE_MAX_ETHPORTS];
154 static unsigned nb_ports; /**< The number of ports specified in command line */
156 /* ethernet addresses of ports */
157 struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
159 /* heads for the main used and free linked lists for the data path. */
160 static struct virtio_net_data_ll *ll_root_used;
161 static struct virtio_net_data_ll *ll_root_free;
164 * Array of data core structures containing information on
165 * individual core linked lists.
167 static struct lcore_info lcore_info[RTE_MAX_LCORE];
169 /* Used for queueing bursts of TX packets. */
173 struct rte_mbuf *m_table[MAX_PKT_BURST];
176 /* TX queue for each data core. */
177 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
179 struct device_statistics dev_statistics[MAX_DEVICES];
182 * Set character device basename.
185 us_vhost_parse_basename(const char *q_arg)
187 /* parse number string */
188 if (strlen(q_arg) >= MAX_BASENAME_SZ)
191 snprintf((char *)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
197 * Parse the portmask provided at run time.
200 parse_portmask(const char *portmask)
205 /* parse hexadecimal string */
206 pm = strtoul(portmask, &end, 16);
207 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
217 * Parse num options at run time.
220 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
225 /* parse unsigned int string */
226 num = strtoul(q_arg, &end, 10);
227 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
230 if (num > max_valid_value)
240 tep_termination_usage(const char *prgname)
242 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
243 " --nb-devices[1-64]: The number of virtIO device\n"
244 " -p PORTMASK: Set mask for ports to be used by application\n"
245 " --rx-retry [0|1]: disable/enable(default) retries on rx."
246 " Enable retry if destintation queue is full\n"
247 " --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX."
248 " This makes effect only if retries on rx enabled\n"
249 " --rx-retry-num [0-N]: the number of retries on rx."
250 " This makes effect only if retries on rx enabled\n"
251 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
252 " --dev-basename: The basename to be used for the character device.\n",
257 * Parse the arguments given in the command line of the application.
260 tep_termination_parse_args(int argc, char **argv)
265 const char *prgname = argv[0];
266 static struct option long_option[] = {
267 {CMD_LINE_OPT_NB_DEVICES, required_argument, NULL, 0},
268 {CMD_LINE_OPT_RX_RETRY, required_argument, NULL, 0},
269 {CMD_LINE_OPT_RX_RETRY_DELAY, required_argument, NULL, 0},
270 {CMD_LINE_OPT_RX_RETRY_NUM, required_argument, NULL, 0},
271 {CMD_LINE_OPT_STATS, required_argument, NULL, 0},
272 {CMD_LINE_OPT_DEV_BASENAME, required_argument, NULL, 0},
276 /* Parse command line */
277 while ((opt = getopt_long(argc, argv, "p:",
278 long_option, &option_index)) != EOF) {
282 enabled_port_mask = parse_portmask(optarg);
283 if (enabled_port_mask == 0) {
284 RTE_LOG(INFO, VHOST_CONFIG,
285 "Invalid portmask\n");
286 tep_termination_usage(prgname);
291 if (!strncmp(long_option[option_index].name,
292 CMD_LINE_OPT_NB_DEVICES,
293 sizeof(CMD_LINE_OPT_NB_DEVICES))) {
294 ret = parse_num_opt(optarg, MAX_DEVICES);
296 RTE_LOG(INFO, VHOST_CONFIG,
297 "Invalid argument for nb-devices [0-%d]\n",
299 tep_termination_usage(prgname);
305 /* Enable/disable retries on RX. */
306 if (!strncmp(long_option[option_index].name,
307 CMD_LINE_OPT_RX_RETRY,
308 sizeof(CMD_LINE_OPT_RX_RETRY))) {
309 ret = parse_num_opt(optarg, 1);
311 RTE_LOG(INFO, VHOST_CONFIG,
312 "Invalid argument for rx-retry [0|1]\n");
313 tep_termination_usage(prgname);
320 /* Specify the retries delay time (in useconds) on RX.*/
321 if (!strncmp(long_option[option_index].name,
322 CMD_LINE_OPT_RX_RETRY_DELAY,
323 sizeof(CMD_LINE_OPT_RX_RETRY_DELAY))) {
324 ret = parse_num_opt(optarg, INT32_MAX);
326 RTE_LOG(INFO, VHOST_CONFIG,
327 "Invalid argument for rx-retry-delay [0-N]\n");
328 tep_termination_usage(prgname);
331 burst_rx_delay_time = ret;
335 /* Specify the retries number on RX. */
336 if (!strncmp(long_option[option_index].name,
337 CMD_LINE_OPT_RX_RETRY_NUM,
338 sizeof(CMD_LINE_OPT_RX_RETRY_NUM))) {
339 ret = parse_num_opt(optarg, INT32_MAX);
341 RTE_LOG(INFO, VHOST_CONFIG,
342 "Invalid argument for rx-retry-num [0-N]\n");
343 tep_termination_usage(prgname);
346 burst_rx_retry_num = ret;
350 /* Enable/disable stats. */
351 if (!strncmp(long_option[option_index].name,
353 sizeof(CMD_LINE_OPT_STATS))) {
354 ret = parse_num_opt(optarg, INT32_MAX);
356 RTE_LOG(INFO, VHOST_CONFIG,
357 "Invalid argument for stats [0..N]\n");
358 tep_termination_usage(prgname);
365 /* Set character device basename. */
366 if (!strncmp(long_option[option_index].name,
367 CMD_LINE_OPT_DEV_BASENAME,
368 sizeof(CMD_LINE_OPT_DEV_BASENAME))) {
369 if (us_vhost_parse_basename(optarg) == -1) {
370 RTE_LOG(INFO, VHOST_CONFIG,
371 "Invalid argument for character "
372 "device basename (Max %d characters)\n",
374 tep_termination_usage(prgname);
381 /* Invalid option - print options. */
383 tep_termination_usage(prgname);
388 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
389 if (enabled_port_mask & (1 << i))
390 ports[nb_ports++] = (uint8_t)i;
393 if ((nb_ports == 0) || (nb_ports > MAX_SUP_PORTS)) {
394 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
395 "but only %u port can be enabled\n", nb_ports,
404 * Update the global var NB_PORTS and array PORTS
405 * according to system ports number and return valid ports number
408 check_ports_num(unsigned max_nb_ports)
410 unsigned valid_nb_ports = nb_ports;
413 if (nb_ports > max_nb_ports) {
414 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) "
415 " exceeds total system port number(%u)\n",
416 nb_ports, max_nb_ports);
417 nb_ports = max_nb_ports;
420 for (portid = 0; portid < nb_ports; portid++) {
421 if (ports[portid] >= max_nb_ports) {
422 RTE_LOG(INFO, VHOST_PORT,
423 "\nSpecified port ID(%u) exceeds max "
424 " system port ID(%u)\n",
425 ports[portid], (max_nb_ports - 1));
426 ports[portid] = INVALID_PORT_ID;
430 return valid_nb_ports;
434 * This function routes the TX packet to the correct interface. This may be a local device
435 * or the physical port.
437 static inline void __attribute__((always_inline))
438 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
440 struct mbuf_table *tx_q;
441 struct rte_mbuf **m_table;
442 unsigned len, ret = 0;
443 const uint16_t lcore_id = rte_lcore_id();
444 struct virtio_net *dev = vdev->dev;
446 LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is external\n",
449 /* Add packet to the port tx queue */
450 tx_q = &lcore_tx_queue[lcore_id];
453 tx_q->m_table[len] = m;
456 dev_statistics[dev->device_fh].tx_total++;
457 dev_statistics[dev->device_fh].tx++;
460 if (unlikely(len == MAX_PKT_BURST)) {
461 m_table = (struct rte_mbuf **)tx_q->m_table;
462 /* Free any buffers not handled by TX and update
465 if (unlikely(ret < len)) {
467 rte_pktmbuf_free(m_table[ret]);
468 } while (++ret < len);
479 * This function is called by each data core. It handles all
480 * RX/TX registered with the core. For TX the specific lcore
481 * linked list is used. For RX, MAC addresses are compared
482 * with all devices in the main linked list.
485 switch_worker(__rte_unused void *arg)
487 struct rte_mempool *mbuf_pool = arg;
488 struct virtio_net *dev = NULL;
489 struct vhost_dev *vdev = NULL;
490 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
491 struct virtio_net_data_ll *dev_ll;
492 struct mbuf_table *tx_q;
493 volatile struct lcore_ll_info *lcore_ll;
494 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
495 / US_PER_S * BURST_TX_DRAIN_US;
496 uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
498 const uint16_t lcore_id = rte_lcore_id();
499 const uint16_t num_cores = (uint16_t)rte_lcore_count();
500 uint16_t rx_count = 0;
504 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
505 lcore_ll = lcore_info[lcore_id].lcore_ll;
508 tx_q = &lcore_tx_queue[lcore_id];
509 for (i = 0; i < num_cores; i++) {
510 if (lcore_ids[i] == lcore_id) {
517 cur_tsc = rte_rdtsc();
519 * TX burst queue drain
521 diff_tsc = cur_tsc - prev_tsc;
522 if (unlikely(diff_tsc > drain_tsc)) {
525 LOG_DEBUG(VHOST_DATA, "TX queue drained after "
526 "timeout with burst size %u\n",
528 if (unlikely(ret < tx_q->len)) {
530 rte_pktmbuf_free(tx_q->m_table[ret]);
531 } while (++ret < tx_q->len);
541 rte_prefetch0(lcore_ll->ll_root_used);
544 * Inform the configuration core that we have exited
545 * the linked list and that no devices are
546 * in use if requested.
548 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
549 lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
554 dev_ll = lcore_ll->ll_root_used;
556 while (dev_ll != NULL) {
560 if (unlikely(vdev->remove)) {
561 dev_ll = dev_ll->next;
562 vdev->ready = DEVICE_SAFE_REMOVE;
565 if (likely(vdev->ready == DEVICE_RX)) {
566 /* Handle guest RX */
567 rx_count = rte_eth_rx_burst(ports[0],
568 vdev->rx_q, pkts_burst, MAX_PKT_BURST);
572 * Retry is enabled and the queue is
573 * full then we wait and retry to
574 * avoid packet loss. Here MAX_PKT_BURST
575 * must be less than virtio queue size
577 if (enable_retry && unlikely(rx_count >
578 rte_vring_available_entries(dev, VIRTIO_RXQ))) {
579 for (retry = 0; retry < burst_rx_retry_num;
581 rte_delay_us(burst_rx_delay_time);
582 if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
589 &dev_statistics[dev->device_fh].rx_total_atomic,
592 &dev_statistics[dev->device_fh].rx_atomic, ret_count);
594 while (likely(rx_count)) {
596 rte_pktmbuf_free(pkts_burst[rx_count]);
602 if (likely(!vdev->remove)) {
604 tx_count = rte_vhost_dequeue_burst(dev,
605 VIRTIO_TXQ, mbuf_pool,
606 pkts_burst, MAX_PKT_BURST);
607 /* If this is the first received packet we need to learn the MAC */
608 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
611 rte_pktmbuf_free(pkts_burst[--tx_count]);
615 virtio_tx_route(vdev, pkts_burst[--tx_count]);
618 /* move to the next device in the list */
619 dev_ll = dev_ll->next;
627 * Add an entry to a used linked list. A free entry must first be found
628 * in the free linked list using get_data_ll_free_entry();
631 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
632 struct virtio_net_data_ll *ll_dev)
634 struct virtio_net_data_ll *ll = *ll_root_addr;
636 /* Set next as NULL and use a compiler barrier to avoid reordering. */
638 rte_compiler_barrier();
640 /* If ll == NULL then this is the first device. */
642 /* Increment to the tail of the linked list. */
643 while (ll->next != NULL)
648 *ll_root_addr = ll_dev;
653 * Remove an entry from a used linked list. The entry must then be added to
654 * the free linked list using put_data_ll_free_entry().
657 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
658 struct virtio_net_data_ll *ll_dev,
659 struct virtio_net_data_ll *ll_dev_last)
661 struct virtio_net_data_ll *ll = *ll_root_addr;
663 if (unlikely((ll == NULL) || (ll_dev == NULL)))
667 *ll_root_addr = ll_dev->next;
669 if (likely(ll_dev_last != NULL))
670 ll_dev_last->next = ll_dev->next;
672 RTE_LOG(ERR, VHOST_CONFIG,
673 "Remove entry form ll failed.\n");
677 * Find and return an entry from the free linked list.
679 static struct virtio_net_data_ll *
680 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
682 struct virtio_net_data_ll *ll_free = *ll_root_addr;
683 struct virtio_net_data_ll *ll_dev;
689 *ll_root_addr = ll_free->next;
695 * Place an entry back on to the free linked list.
698 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
699 struct virtio_net_data_ll *ll_dev)
701 struct virtio_net_data_ll *ll_free = *ll_root_addr;
706 ll_dev->next = ll_free;
707 *ll_root_addr = ll_dev;
711 * Creates a linked list of a given size.
713 static struct virtio_net_data_ll *
714 alloc_data_ll(uint32_t size)
716 struct virtio_net_data_ll *ll_new;
719 /* Malloc and then chain the linked list. */
720 ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
721 if (ll_new == NULL) {
722 RTE_LOG(ERR, VHOST_CONFIG,
723 "Failed to allocate memory for ll_new.\n");
727 for (i = 0; i < size - 1; i++) {
728 ll_new[i].vdev = NULL;
729 ll_new[i].next = &ll_new[i+1];
731 ll_new[i].next = NULL;
737 * Create the main linked list along with each individual cores
738 * linked list. A used and a free list are created to manage entries.
745 RTE_LCORE_FOREACH_SLAVE(lcore) {
746 lcore_info[lcore].lcore_ll =
747 malloc(sizeof(struct lcore_ll_info));
748 if (lcore_info[lcore].lcore_ll == NULL) {
749 RTE_LOG(ERR, VHOST_CONFIG,
750 "Failed to allocate memory for lcore_ll.\n");
754 lcore_info[lcore].lcore_ll->device_num = 0;
755 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
756 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
757 if (nb_devices % nb_switching_cores)
758 lcore_info[lcore].lcore_ll->ll_root_free =
759 alloc_data_ll((nb_devices / nb_switching_cores)
762 lcore_info[lcore].lcore_ll->ll_root_free =
763 alloc_data_ll(nb_devices / nb_switching_cores);
766 /* Allocate devices up to a maximum of MAX_DEVICES. */
767 ll_root_free = alloc_data_ll(MIN((nb_devices), MAX_DEVICES));
773 * Remove a device from the specific data core linked list and
774 * from the main linked list. Synchonization occurs through the use
775 * of the lcore dev_removal_flag. Device is made volatile here
776 * to avoid re-ordering of dev->remove=1 which can cause an infinite
777 * loop in the rte_pause loop.
780 destroy_device(volatile struct virtio_net *dev)
782 struct virtio_net_data_ll *ll_lcore_dev_cur;
783 struct virtio_net_data_ll *ll_main_dev_cur;
784 struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
785 struct virtio_net_data_ll *ll_main_dev_last = NULL;
786 struct vhost_dev *vdev;
789 dev->flags &= ~VIRTIO_DEV_RUNNING;
791 vdev = (struct vhost_dev *)dev->priv;
793 /* set the remove flag. */
795 while (vdev->ready != DEVICE_SAFE_REMOVE)
798 /* Search for entry to be removed from lcore ll */
799 ll_lcore_dev_cur = lcore_info[vdev->coreid].lcore_ll->ll_root_used;
800 while (ll_lcore_dev_cur != NULL) {
801 if (ll_lcore_dev_cur->vdev == vdev) {
804 ll_lcore_dev_last = ll_lcore_dev_cur;
805 ll_lcore_dev_cur = ll_lcore_dev_cur->next;
809 if (ll_lcore_dev_cur == NULL) {
810 RTE_LOG(ERR, VHOST_CONFIG,
811 "(%"PRIu64") Failed to find the dev to be destroy.\n",
816 /* Search for entry to be removed from main ll */
817 ll_main_dev_cur = ll_root_used;
818 ll_main_dev_last = NULL;
819 while (ll_main_dev_cur != NULL) {
820 if (ll_main_dev_cur->vdev == vdev) {
823 ll_main_dev_last = ll_main_dev_cur;
824 ll_main_dev_cur = ll_main_dev_cur->next;
828 /* Remove entries from the lcore and main ll. */
829 rm_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
830 ll_lcore_dev_cur, ll_lcore_dev_last);
831 rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
833 /* Set the dev_removal_flag on each lcore. */
834 RTE_LCORE_FOREACH_SLAVE(lcore) {
835 lcore_info[lcore].lcore_ll->dev_removal_flag =
840 * Once each core has set the dev_removal_flag to
841 * ACK_DEV_REMOVAL we can be sure that they can no longer access
842 * the device removed from the linked lists and that the devices
843 * are no longer in use.
845 RTE_LCORE_FOREACH_SLAVE(lcore) {
846 while (lcore_info[lcore].lcore_ll->dev_removal_flag
851 /* Add the entries back to the lcore and main free ll.*/
852 put_data_ll_free_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_free,
854 put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
856 /* Decrement number of device on the lcore. */
857 lcore_info[vdev->coreid].lcore_ll->device_num--;
859 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been removed "
860 "from data core\n", dev->device_fh);
867 * A new device is added to a data core. First the device is added
868 * to the main linked list and the allocated to a specific data core.
871 new_device(struct virtio_net *dev)
873 struct virtio_net_data_ll *ll_dev;
874 int lcore, core_add = 0;
875 uint32_t device_num_min = nb_devices;
876 struct vhost_dev *vdev;
878 vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
880 RTE_LOG(INFO, VHOST_DATA,
881 "(%"PRIu64") Couldn't allocate memory for vhost dev\n",
887 /* Add device to main ll */
888 ll_dev = get_data_ll_free_entry(&ll_root_free);
889 if (ll_dev == NULL) {
890 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in"
891 " linked list Device limit of %d devices per core"
892 " has been reached\n", dev->device_fh, nb_devices);
893 if (vdev->regions_hpa)
894 rte_free(vdev->regions_hpa);
899 add_data_ll_entry(&ll_root_used, ll_dev);
900 vdev->rx_q = dev->device_fh;
902 /* reset ready flag */
903 vdev->ready = DEVICE_MAC_LEARNING;
906 /* Find a suitable lcore to add the device. */
907 RTE_LCORE_FOREACH_SLAVE(lcore) {
908 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
909 device_num_min = lcore_info[lcore].lcore_ll->device_num;
913 /* Add device to lcore ll */
914 ll_dev = get_data_ll_free_entry(&lcore_info[core_add].lcore_ll->ll_root_free);
915 if (ll_dev == NULL) {
916 RTE_LOG(INFO, VHOST_DATA,
917 "(%"PRIu64") Failed to add device to data core\n",
919 vdev->ready = DEVICE_SAFE_REMOVE;
921 rte_free(vdev->regions_hpa);
926 vdev->coreid = core_add;
928 add_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
931 /* Initialize device stats */
932 memset(&dev_statistics[dev->device_fh], 0,
933 sizeof(struct device_statistics));
935 /* Disable notifications. */
936 rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
937 rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
938 lcore_info[vdev->coreid].lcore_ll->device_num++;
939 dev->flags |= VIRTIO_DEV_RUNNING;
941 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n",
942 dev->device_fh, vdev->coreid);
948 * These callback allow devices to be added to the data core when configuration
949 * has been fully complete.
951 static const struct virtio_net_device_ops virtio_net_device_ops = {
952 .new_device = new_device,
953 .destroy_device = destroy_device,
957 * This is a thread will wake up after a period to print stats if the user has
963 struct virtio_net_data_ll *dev_ll;
964 uint64_t tx_dropped, rx_dropped;
965 uint64_t tx, tx_total, rx, rx_total;
967 const char clr[] = { 27, '[', '2', 'J', '\0' };
968 const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
973 /* Clear screen and move to top left */
974 printf("%s%s", clr, top_left);
976 printf("\nDevice statistics ================================");
978 dev_ll = ll_root_used;
979 while (dev_ll != NULL) {
980 device_fh = (uint32_t)dev_ll->vdev->dev->device_fh;
981 tx_total = dev_statistics[device_fh].tx_total;
982 tx = dev_statistics[device_fh].tx;
983 tx_dropped = tx_total - tx;
985 rx_total = rte_atomic64_read(
986 &dev_statistics[device_fh].rx_total_atomic);
987 rx = rte_atomic64_read(
988 &dev_statistics[device_fh].rx_atomic);
989 rx_dropped = rx_total - rx;
991 printf("\nStatistics for device %"PRIu32" ----------"
992 "\nTX total: %"PRIu64""
993 "\nTX dropped: %"PRIu64""
994 "\nTX successful: %"PRIu64""
995 "\nRX total: %"PRIu64""
996 "\nRX dropped: %"PRIu64""
997 "\nRX successful: %"PRIu64"",
1006 dev_ll = dev_ll->next;
1008 printf("\n================================================\n");
1013 * Main function, does initialisation and calls the per-lcore functions. The CUSE
1014 * device is also registered here to handle the IOCTLs.
1017 main(int argc, char *argv[])
1019 struct rte_mempool *mbuf_pool = NULL;
1020 unsigned lcore_id, core_id = 0;
1021 unsigned nb_ports, valid_nb_ports;
1025 static pthread_t tid;
1028 ret = rte_eal_init(argc, argv);
1030 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1034 /* parse app arguments */
1035 ret = tep_termination_parse_args(argc, argv);
1037 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1039 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
1040 if (rte_lcore_is_enabled(lcore_id))
1041 lcore_ids[core_id++] = lcore_id;
1043 /* set the number of swithcing cores available */
1044 nb_switching_cores = rte_lcore_count()-1;
1046 /* Get the number of physical ports. */
1047 nb_ports = rte_eth_dev_count();
1048 if (nb_ports > RTE_MAX_ETHPORTS)
1049 nb_ports = RTE_MAX_ETHPORTS;
1052 * Update the global var NB_PORTS and global array PORTS
1053 * and get value of var VALID_NB_PORTS according to system ports number
1055 valid_nb_ports = check_ports_num(nb_ports);
1057 if ((valid_nb_ports == 0) || (valid_nb_ports > MAX_SUP_PORTS)) {
1058 rte_exit(EXIT_FAILURE, "Current enabled port number is %u,"
1059 "but only %u port can be enabled\n", nb_ports,
1063 /* Create the mbuf pool. */
1064 mbuf_pool = rte_mempool_create(
1068 MBUF_SIZE, MBUF_CACHE_SIZE,
1069 sizeof(struct rte_pktmbuf_pool_private),
1070 rte_pktmbuf_pool_init, NULL,
1071 rte_pktmbuf_init, NULL,
1072 rte_socket_id(), 0);
1073 if (mbuf_pool == NULL)
1074 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1076 for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
1077 vpool_array[queue_id].pool = mbuf_pool;
1079 /* Set log level. */
1080 rte_set_log_level(LOG_LEVEL);
1082 /* initialize all ports */
1083 for (portid = 0; portid < nb_ports; portid++) {
1084 /* skip ports that are not enabled */
1085 if ((enabled_port_mask & (1 << portid)) == 0) {
1086 RTE_LOG(INFO, VHOST_PORT,
1087 "Skipping disabled port %d\n", portid);
1092 /* Initialise all linked lists. */
1093 if (init_data_ll() == -1)
1094 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1096 /* Initialize device stats */
1097 memset(&dev_statistics, 0, sizeof(dev_statistics));
1099 /* Enable stats if the user option is set. */
1101 pthread_create(&tid, NULL, (void *)print_stats, NULL);
1103 /* Launch all data cores. */
1104 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1105 rte_eal_remote_launch(switch_worker,
1106 mbuf_pool, lcore_id);
1109 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1111 /* Register CUSE device to handle IOCTLs. */
1112 ret = rte_vhost_driver_register((char *)&dev_basename);
1114 rte_exit(EXIT_FAILURE, "CUSE device setup failure.\n");
1116 rte_vhost_driver_callback_register(&virtio_net_device_ops);
1118 /* Start CUSE session. */
1119 rte_vhost_driver_session_start();