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_UDP_PORT "udp-port"
116 #define CMD_LINE_OPT_TX_CHECKSUM "tx-checksum"
117 #define CMD_LINE_OPT_FILTER_TYPE "filter-type"
118 #define CMD_LINE_OPT_RX_RETRY "rx-retry"
119 #define CMD_LINE_OPT_RX_RETRY_DELAY "rx-retry-delay"
120 #define CMD_LINE_OPT_RX_RETRY_NUM "rx-retry-num"
121 #define CMD_LINE_OPT_STATS "stats"
122 #define CMD_LINE_OPT_DEV_BASENAME "dev-basename"
124 /* mask of enabled ports */
125 static uint32_t enabled_port_mask;
127 /*Number of switching cores enabled*/
128 static uint32_t nb_switching_cores;
130 /* number of devices/queues to support*/
131 uint16_t nb_devices = 2;
133 /* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
134 #define MAX_RING_DESC 4096
137 struct rte_mempool *pool;
138 struct rte_ring *ring;
140 } vpool_array[MAX_QUEUES+MAX_QUEUES];
142 /* UDP tunneling port */
143 uint16_t udp_port = 4789;
145 /* enable/disable inner TX checksum */
146 uint8_t tx_checksum = 0;
148 /* RX filter type for tunneling packet */
149 uint8_t filter_idx = 1;
151 /* overlay packet operation */
152 struct ol_switch_ops overlay_options = {
153 .port_configure = vxlan_port_init,
154 .tunnel_setup = vxlan_link,
155 .tunnel_destroy = vxlan_unlink,
156 .tx_handle = vxlan_tx_pkts,
157 .rx_handle = vxlan_rx_pkts,
158 .param_handle = NULL,
162 uint32_t enable_stats = 0;
163 /* Enable retries on RX. */
164 static uint32_t enable_retry = 1;
165 /* Specify timeout (in useconds) between retries on RX. */
166 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
167 /* Specify the number of retries on RX. */
168 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
170 /* Character device basename. Can be set by user. */
171 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
173 static unsigned lcore_ids[RTE_MAX_LCORE];
174 uint8_t ports[RTE_MAX_ETHPORTS];
176 static unsigned nb_ports; /**< The number of ports specified in command line */
178 /* ethernet addresses of ports */
179 struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
181 /* heads for the main used and free linked lists for the data path. */
182 static struct virtio_net_data_ll *ll_root_used;
183 static struct virtio_net_data_ll *ll_root_free;
186 * Array of data core structures containing information on
187 * individual core linked lists.
189 static struct lcore_info lcore_info[RTE_MAX_LCORE];
191 /* Used for queueing bursts of TX packets. */
195 struct rte_mbuf *m_table[MAX_PKT_BURST];
198 /* TX queue for each data core. */
199 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
201 struct device_statistics dev_statistics[MAX_DEVICES];
204 * Set character device basename.
207 us_vhost_parse_basename(const char *q_arg)
209 /* parse number string */
210 if (strlen(q_arg) >= MAX_BASENAME_SZ)
213 snprintf((char *)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
219 * Parse the portmask provided at run time.
222 parse_portmask(const char *portmask)
227 /* parse hexadecimal string */
228 pm = strtoul(portmask, &end, 16);
229 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
239 * Parse num options at run time.
242 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
247 /* parse unsigned int string */
248 num = strtoul(q_arg, &end, 10);
249 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
252 if (num > max_valid_value)
262 tep_termination_usage(const char *prgname)
264 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
265 " --udp-port: UDP destination port for VXLAN packet\n"
266 " --nb-devices[1-64]: The number of virtIO device\n"
267 " --tx-checksum [0|1]: inner Tx checksum offload\n"
268 " --filter-type[1-3]: filter type for tunneling packet\n"
269 " 1: Inner MAC and tenent ID\n"
270 " 2: Inner MAC and VLAN, and tenent ID\n"
271 " 3: Outer MAC, Inner MAC and tenent ID\n"
272 " -p PORTMASK: Set mask for ports to be used by application\n"
273 " --rx-retry [0|1]: disable/enable(default) retries on rx."
274 " Enable retry if destintation queue is full\n"
275 " --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX."
276 " This makes effect only if retries on rx enabled\n"
277 " --rx-retry-num [0-N]: the number of retries on rx."
278 " This makes effect only if retries on rx enabled\n"
279 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
280 " --dev-basename: The basename to be used for the character device.\n",
285 * Parse the arguments given in the command line of the application.
288 tep_termination_parse_args(int argc, char **argv)
293 const char *prgname = argv[0];
294 static struct option long_option[] = {
295 {CMD_LINE_OPT_NB_DEVICES, required_argument, NULL, 0},
296 {CMD_LINE_OPT_UDP_PORT, required_argument, NULL, 0},
297 {CMD_LINE_OPT_TX_CHECKSUM, required_argument, NULL, 0},
298 {CMD_LINE_OPT_FILTER_TYPE, required_argument, NULL, 0},
299 {CMD_LINE_OPT_RX_RETRY, required_argument, NULL, 0},
300 {CMD_LINE_OPT_RX_RETRY_DELAY, required_argument, NULL, 0},
301 {CMD_LINE_OPT_RX_RETRY_NUM, required_argument, NULL, 0},
302 {CMD_LINE_OPT_STATS, required_argument, NULL, 0},
303 {CMD_LINE_OPT_DEV_BASENAME, required_argument, NULL, 0},
307 /* Parse command line */
308 while ((opt = getopt_long(argc, argv, "p:",
309 long_option, &option_index)) != EOF) {
313 enabled_port_mask = parse_portmask(optarg);
314 if (enabled_port_mask == 0) {
315 RTE_LOG(INFO, VHOST_CONFIG,
316 "Invalid portmask\n");
317 tep_termination_usage(prgname);
322 if (!strncmp(long_option[option_index].name,
323 CMD_LINE_OPT_NB_DEVICES,
324 sizeof(CMD_LINE_OPT_NB_DEVICES))) {
325 ret = parse_num_opt(optarg, MAX_DEVICES);
327 RTE_LOG(INFO, VHOST_CONFIG,
328 "Invalid argument for nb-devices [0-%d]\n",
330 tep_termination_usage(prgname);
336 /* Enable/disable retries on RX. */
337 if (!strncmp(long_option[option_index].name,
338 CMD_LINE_OPT_RX_RETRY,
339 sizeof(CMD_LINE_OPT_RX_RETRY))) {
340 ret = parse_num_opt(optarg, 1);
342 RTE_LOG(INFO, VHOST_CONFIG,
343 "Invalid argument for rx-retry [0|1]\n");
344 tep_termination_usage(prgname);
350 if (!strncmp(long_option[option_index].name,
351 CMD_LINE_OPT_UDP_PORT,
352 sizeof(CMD_LINE_OPT_UDP_PORT))) {
353 ret = parse_num_opt(optarg, INT16_MAX);
355 RTE_LOG(INFO, VHOST_CONFIG,
356 "Invalid argument for UDP port [0-N]\n");
357 tep_termination_usage(prgname);
363 /* Specify the retries delay time (in useconds) on RX.*/
364 if (!strncmp(long_option[option_index].name,
365 CMD_LINE_OPT_RX_RETRY_DELAY,
366 sizeof(CMD_LINE_OPT_RX_RETRY_DELAY))) {
367 ret = parse_num_opt(optarg, INT32_MAX);
369 RTE_LOG(INFO, VHOST_CONFIG,
370 "Invalid argument for rx-retry-delay [0-N]\n");
371 tep_termination_usage(prgname);
374 burst_rx_delay_time = ret;
377 /* Specify the retries number on RX. */
378 if (!strncmp(long_option[option_index].name,
379 CMD_LINE_OPT_RX_RETRY_NUM,
380 sizeof(CMD_LINE_OPT_RX_RETRY_NUM))) {
381 ret = parse_num_opt(optarg, INT32_MAX);
383 RTE_LOG(INFO, VHOST_CONFIG,
384 "Invalid argument for rx-retry-num [0-N]\n");
385 tep_termination_usage(prgname);
388 burst_rx_retry_num = ret;
391 if (!strncmp(long_option[option_index].name,
392 CMD_LINE_OPT_TX_CHECKSUM,
393 sizeof(CMD_LINE_OPT_TX_CHECKSUM))) {
394 ret = parse_num_opt(optarg, 1);
396 RTE_LOG(INFO, VHOST_CONFIG,
397 "Invalid argument for tx-checksum [0|1]\n");
398 tep_termination_usage(prgname);
404 if (!strncmp(long_option[option_index].name,
405 CMD_LINE_OPT_FILTER_TYPE,
406 sizeof(CMD_LINE_OPT_FILTER_TYPE))) {
407 ret = parse_num_opt(optarg, 3);
408 if ((ret == -1) || (ret == 0)) {
409 RTE_LOG(INFO, VHOST_CONFIG,
410 "Invalid argument for filter type [1-3]\n");
411 tep_termination_usage(prgname);
414 filter_idx = ret - 1;
417 /* Enable/disable stats. */
418 if (!strncmp(long_option[option_index].name,
420 sizeof(CMD_LINE_OPT_STATS))) {
421 ret = parse_num_opt(optarg, INT32_MAX);
423 RTE_LOG(INFO, VHOST_CONFIG,
424 "Invalid argument for stats [0..N]\n");
425 tep_termination_usage(prgname);
431 /* Set character device basename. */
432 if (!strncmp(long_option[option_index].name,
433 CMD_LINE_OPT_DEV_BASENAME,
434 sizeof(CMD_LINE_OPT_DEV_BASENAME))) {
435 if (us_vhost_parse_basename(optarg) == -1) {
436 RTE_LOG(INFO, VHOST_CONFIG,
437 "Invalid argument for character "
438 "device basename (Max %d characters)\n",
440 tep_termination_usage(prgname);
447 /* Invalid option - print options. */
449 tep_termination_usage(prgname);
454 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
455 if (enabled_port_mask & (1 << i))
456 ports[nb_ports++] = (uint8_t)i;
459 if ((nb_ports == 0) || (nb_ports > MAX_SUP_PORTS)) {
460 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
461 "but only %u port can be enabled\n", nb_ports,
470 * Update the global var NB_PORTS and array PORTS
471 * according to system ports number and return valid ports number
474 check_ports_num(unsigned max_nb_ports)
476 unsigned valid_nb_ports = nb_ports;
479 if (nb_ports > max_nb_ports) {
480 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) "
481 " exceeds total system port number(%u)\n",
482 nb_ports, max_nb_ports);
483 nb_ports = max_nb_ports;
486 for (portid = 0; portid < nb_ports; portid++) {
487 if (ports[portid] >= max_nb_ports) {
488 RTE_LOG(INFO, VHOST_PORT,
489 "\nSpecified port ID(%u) exceeds max "
490 " system port ID(%u)\n",
491 ports[portid], (max_nb_ports - 1));
492 ports[portid] = INVALID_PORT_ID;
496 return valid_nb_ports;
500 * This function routes the TX packet to the correct interface. This may be a local device
501 * or the physical port.
503 static inline void __attribute__((always_inline))
504 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
506 struct mbuf_table *tx_q;
507 struct rte_mbuf **m_table;
508 unsigned len, ret = 0;
509 const uint16_t lcore_id = rte_lcore_id();
510 struct virtio_net *dev = vdev->dev;
512 LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is external\n",
515 /* Add packet to the port tx queue */
516 tx_q = &lcore_tx_queue[lcore_id];
519 tx_q->m_table[len] = m;
522 dev_statistics[dev->device_fh].tx_total++;
523 dev_statistics[dev->device_fh].tx++;
526 if (unlikely(len == MAX_PKT_BURST)) {
527 m_table = (struct rte_mbuf **)tx_q->m_table;
528 ret = overlay_options.tx_handle(ports[0],
529 (uint16_t)tx_q->txq_id, m_table,
530 (uint16_t)tx_q->len);
532 /* Free any buffers not handled by TX and update
535 if (unlikely(ret < len)) {
537 rte_pktmbuf_free(m_table[ret]);
538 } while (++ret < len);
549 * This function is called by each data core. It handles all
550 * RX/TX registered with the core. For TX the specific lcore
551 * linked list is used. For RX, MAC addresses are compared
552 * with all devices in the main linked list.
555 switch_worker(__rte_unused void *arg)
557 struct rte_mempool *mbuf_pool = arg;
558 struct virtio_net *dev = NULL;
559 struct vhost_dev *vdev = NULL;
560 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
561 struct virtio_net_data_ll *dev_ll;
562 struct mbuf_table *tx_q;
563 volatile struct lcore_ll_info *lcore_ll;
564 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
565 / US_PER_S * BURST_TX_DRAIN_US;
566 uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
568 const uint16_t lcore_id = rte_lcore_id();
569 const uint16_t num_cores = (uint16_t)rte_lcore_count();
570 uint16_t rx_count = 0;
574 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
575 lcore_ll = lcore_info[lcore_id].lcore_ll;
578 tx_q = &lcore_tx_queue[lcore_id];
579 for (i = 0; i < num_cores; i++) {
580 if (lcore_ids[i] == lcore_id) {
587 cur_tsc = rte_rdtsc();
589 * TX burst queue drain
591 diff_tsc = cur_tsc - prev_tsc;
592 if (unlikely(diff_tsc > drain_tsc)) {
595 LOG_DEBUG(VHOST_DATA, "TX queue drained after "
596 "timeout with burst size %u\n",
598 ret = overlay_options.tx_handle(ports[0],
599 (uint16_t)tx_q->txq_id,
600 (struct rte_mbuf **)tx_q->m_table,
601 (uint16_t)tx_q->len);
602 if (unlikely(ret < tx_q->len)) {
604 rte_pktmbuf_free(tx_q->m_table[ret]);
605 } while (++ret < tx_q->len);
615 rte_prefetch0(lcore_ll->ll_root_used);
618 * Inform the configuration core that we have exited
619 * the linked list and that no devices are
620 * in use if requested.
622 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
623 lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
628 dev_ll = lcore_ll->ll_root_used;
630 while (dev_ll != NULL) {
634 if (unlikely(vdev->remove)) {
635 dev_ll = dev_ll->next;
636 overlay_options.tunnel_destroy(vdev);
637 vdev->ready = DEVICE_SAFE_REMOVE;
640 if (likely(vdev->ready == DEVICE_RX)) {
641 /* Handle guest RX */
642 rx_count = rte_eth_rx_burst(ports[0],
643 vdev->rx_q, pkts_burst, MAX_PKT_BURST);
647 * Retry is enabled and the queue is
648 * full then we wait and retry to
649 * avoid packet loss. Here MAX_PKT_BURST
650 * must be less than virtio queue size
652 if (enable_retry && unlikely(rx_count >
653 rte_vring_available_entries(dev, VIRTIO_RXQ))) {
654 for (retry = 0; retry < burst_rx_retry_num;
656 rte_delay_us(burst_rx_delay_time);
657 if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
662 ret_count = overlay_options.rx_handle(dev, pkts_burst, rx_count);
665 &dev_statistics[dev->device_fh].rx_total_atomic,
668 &dev_statistics[dev->device_fh].rx_atomic, ret_count);
670 while (likely(rx_count)) {
672 rte_pktmbuf_free(pkts_burst[rx_count]);
678 if (likely(!vdev->remove)) {
680 tx_count = rte_vhost_dequeue_burst(dev,
681 VIRTIO_TXQ, mbuf_pool,
682 pkts_burst, MAX_PKT_BURST);
683 /* If this is the first received packet we need to learn the MAC */
684 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
686 (overlay_options.tunnel_setup(vdev, pkts_burst[0]) == -1)) {
688 rte_pktmbuf_free(pkts_burst[--tx_count]);
692 virtio_tx_route(vdev, pkts_burst[--tx_count]);
695 /* move to the next device in the list */
696 dev_ll = dev_ll->next;
704 * Add an entry to a used linked list. A free entry must first be found
705 * in the free linked list using get_data_ll_free_entry();
708 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
709 struct virtio_net_data_ll *ll_dev)
711 struct virtio_net_data_ll *ll = *ll_root_addr;
713 /* Set next as NULL and use a compiler barrier to avoid reordering. */
715 rte_compiler_barrier();
717 /* If ll == NULL then this is the first device. */
719 /* Increment to the tail of the linked list. */
720 while (ll->next != NULL)
725 *ll_root_addr = ll_dev;
730 * Remove an entry from a used linked list. The entry must then be added to
731 * the free linked list using put_data_ll_free_entry().
734 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
735 struct virtio_net_data_ll *ll_dev,
736 struct virtio_net_data_ll *ll_dev_last)
738 struct virtio_net_data_ll *ll = *ll_root_addr;
740 if (unlikely((ll == NULL) || (ll_dev == NULL)))
744 *ll_root_addr = ll_dev->next;
746 if (likely(ll_dev_last != NULL))
747 ll_dev_last->next = ll_dev->next;
749 RTE_LOG(ERR, VHOST_CONFIG,
750 "Remove entry form ll failed.\n");
754 * Find and return an entry from the free linked list.
756 static struct virtio_net_data_ll *
757 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
759 struct virtio_net_data_ll *ll_free = *ll_root_addr;
760 struct virtio_net_data_ll *ll_dev;
766 *ll_root_addr = ll_free->next;
772 * Place an entry back on to the free linked list.
775 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
776 struct virtio_net_data_ll *ll_dev)
778 struct virtio_net_data_ll *ll_free = *ll_root_addr;
783 ll_dev->next = ll_free;
784 *ll_root_addr = ll_dev;
788 * Creates a linked list of a given size.
790 static struct virtio_net_data_ll *
791 alloc_data_ll(uint32_t size)
793 struct virtio_net_data_ll *ll_new;
796 /* Malloc and then chain the linked list. */
797 ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
798 if (ll_new == NULL) {
799 RTE_LOG(ERR, VHOST_CONFIG,
800 "Failed to allocate memory for ll_new.\n");
804 for (i = 0; i < size - 1; i++) {
805 ll_new[i].vdev = NULL;
806 ll_new[i].next = &ll_new[i+1];
808 ll_new[i].next = NULL;
814 * Create the main linked list along with each individual cores
815 * linked list. A used and a free list are created to manage entries.
822 RTE_LCORE_FOREACH_SLAVE(lcore) {
823 lcore_info[lcore].lcore_ll =
824 malloc(sizeof(struct lcore_ll_info));
825 if (lcore_info[lcore].lcore_ll == NULL) {
826 RTE_LOG(ERR, VHOST_CONFIG,
827 "Failed to allocate memory for lcore_ll.\n");
831 lcore_info[lcore].lcore_ll->device_num = 0;
832 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
833 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
834 if (nb_devices % nb_switching_cores)
835 lcore_info[lcore].lcore_ll->ll_root_free =
836 alloc_data_ll((nb_devices / nb_switching_cores)
839 lcore_info[lcore].lcore_ll->ll_root_free =
840 alloc_data_ll(nb_devices / nb_switching_cores);
843 /* Allocate devices up to a maximum of MAX_DEVICES. */
844 ll_root_free = alloc_data_ll(MIN((nb_devices), MAX_DEVICES));
850 * Remove a device from the specific data core linked list and
851 * from the main linked list. Synchonization occurs through the use
852 * of the lcore dev_removal_flag. Device is made volatile here
853 * to avoid re-ordering of dev->remove=1 which can cause an infinite
854 * loop in the rte_pause loop.
857 destroy_device(volatile struct virtio_net *dev)
859 struct virtio_net_data_ll *ll_lcore_dev_cur;
860 struct virtio_net_data_ll *ll_main_dev_cur;
861 struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
862 struct virtio_net_data_ll *ll_main_dev_last = NULL;
863 struct vhost_dev *vdev;
866 dev->flags &= ~VIRTIO_DEV_RUNNING;
868 vdev = (struct vhost_dev *)dev->priv;
870 /* set the remove flag. */
872 while (vdev->ready != DEVICE_SAFE_REMOVE)
875 /* Search for entry to be removed from lcore ll */
876 ll_lcore_dev_cur = lcore_info[vdev->coreid].lcore_ll->ll_root_used;
877 while (ll_lcore_dev_cur != NULL) {
878 if (ll_lcore_dev_cur->vdev == vdev) {
881 ll_lcore_dev_last = ll_lcore_dev_cur;
882 ll_lcore_dev_cur = ll_lcore_dev_cur->next;
886 if (ll_lcore_dev_cur == NULL) {
887 RTE_LOG(ERR, VHOST_CONFIG,
888 "(%"PRIu64") Failed to find the dev to be destroy.\n",
893 /* Search for entry to be removed from main ll */
894 ll_main_dev_cur = ll_root_used;
895 ll_main_dev_last = NULL;
896 while (ll_main_dev_cur != NULL) {
897 if (ll_main_dev_cur->vdev == vdev) {
900 ll_main_dev_last = ll_main_dev_cur;
901 ll_main_dev_cur = ll_main_dev_cur->next;
905 /* Remove entries from the lcore and main ll. */
906 rm_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
907 ll_lcore_dev_cur, ll_lcore_dev_last);
908 rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
910 /* Set the dev_removal_flag on each lcore. */
911 RTE_LCORE_FOREACH_SLAVE(lcore) {
912 lcore_info[lcore].lcore_ll->dev_removal_flag =
917 * Once each core has set the dev_removal_flag to
918 * ACK_DEV_REMOVAL we can be sure that they can no longer access
919 * the device removed from the linked lists and that the devices
920 * are no longer in use.
922 RTE_LCORE_FOREACH_SLAVE(lcore) {
923 while (lcore_info[lcore].lcore_ll->dev_removal_flag
928 /* Add the entries back to the lcore and main free ll.*/
929 put_data_ll_free_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_free,
931 put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
933 /* Decrement number of device on the lcore. */
934 lcore_info[vdev->coreid].lcore_ll->device_num--;
936 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been removed "
937 "from data core\n", dev->device_fh);
944 * A new device is added to a data core. First the device is added
945 * to the main linked list and the allocated to a specific data core.
948 new_device(struct virtio_net *dev)
950 struct virtio_net_data_ll *ll_dev;
951 int lcore, core_add = 0;
952 uint32_t device_num_min = nb_devices;
953 struct vhost_dev *vdev;
955 vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
957 RTE_LOG(INFO, VHOST_DATA,
958 "(%"PRIu64") Couldn't allocate memory for vhost dev\n",
964 /* Add device to main ll */
965 ll_dev = get_data_ll_free_entry(&ll_root_free);
966 if (ll_dev == NULL) {
967 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in"
968 " linked list Device limit of %d devices per core"
969 " has been reached\n", dev->device_fh, nb_devices);
970 if (vdev->regions_hpa)
971 rte_free(vdev->regions_hpa);
976 add_data_ll_entry(&ll_root_used, ll_dev);
977 vdev->rx_q = dev->device_fh;
979 /* reset ready flag */
980 vdev->ready = DEVICE_MAC_LEARNING;
983 /* Find a suitable lcore to add the device. */
984 RTE_LCORE_FOREACH_SLAVE(lcore) {
985 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
986 device_num_min = lcore_info[lcore].lcore_ll->device_num;
990 /* Add device to lcore ll */
991 ll_dev = get_data_ll_free_entry(&lcore_info[core_add].lcore_ll->ll_root_free);
992 if (ll_dev == NULL) {
993 RTE_LOG(INFO, VHOST_DATA,
994 "(%"PRIu64") Failed to add device to data core\n",
996 vdev->ready = DEVICE_SAFE_REMOVE;
998 rte_free(vdev->regions_hpa);
1002 ll_dev->vdev = vdev;
1003 vdev->coreid = core_add;
1005 add_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
1008 /* Initialize device stats */
1009 memset(&dev_statistics[dev->device_fh], 0,
1010 sizeof(struct device_statistics));
1012 /* Disable notifications. */
1013 rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
1014 rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
1015 lcore_info[vdev->coreid].lcore_ll->device_num++;
1016 dev->flags |= VIRTIO_DEV_RUNNING;
1018 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n",
1019 dev->device_fh, vdev->coreid);
1025 * These callback allow devices to be added to the data core when configuration
1026 * has been fully complete.
1028 static const struct virtio_net_device_ops virtio_net_device_ops = {
1029 .new_device = new_device,
1030 .destroy_device = destroy_device,
1034 * This is a thread will wake up after a period to print stats if the user has
1040 struct virtio_net_data_ll *dev_ll;
1041 uint64_t tx_dropped, rx_dropped;
1042 uint64_t tx, tx_total, rx, rx_total;
1044 const char clr[] = { 27, '[', '2', 'J', '\0' };
1045 const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
1048 sleep(enable_stats);
1050 /* Clear screen and move to top left */
1051 printf("%s%s", clr, top_left);
1053 printf("\nDevice statistics ================================");
1055 dev_ll = ll_root_used;
1056 while (dev_ll != NULL) {
1057 device_fh = (uint32_t)dev_ll->vdev->dev->device_fh;
1058 tx_total = dev_statistics[device_fh].tx_total;
1059 tx = dev_statistics[device_fh].tx;
1060 tx_dropped = tx_total - tx;
1062 rx_total = rte_atomic64_read(
1063 &dev_statistics[device_fh].rx_total_atomic);
1064 rx = rte_atomic64_read(
1065 &dev_statistics[device_fh].rx_atomic);
1066 rx_dropped = rx_total - rx;
1068 printf("\nStatistics for device %"PRIu32" ----------"
1069 "\nTX total: %"PRIu64""
1070 "\nTX dropped: %"PRIu64""
1071 "\nTX successful: %"PRIu64""
1072 "\nRX total: %"PRIu64""
1073 "\nRX dropped: %"PRIu64""
1074 "\nRX successful: %"PRIu64"",
1083 dev_ll = dev_ll->next;
1085 printf("\n================================================\n");
1090 * Main function, does initialisation and calls the per-lcore functions. The CUSE
1091 * device is also registered here to handle the IOCTLs.
1094 main(int argc, char *argv[])
1096 struct rte_mempool *mbuf_pool = NULL;
1097 unsigned lcore_id, core_id = 0;
1098 unsigned nb_ports, valid_nb_ports;
1102 static pthread_t tid;
1105 ret = rte_eal_init(argc, argv);
1107 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1111 /* parse app arguments */
1112 ret = tep_termination_parse_args(argc, argv);
1114 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1116 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
1117 if (rte_lcore_is_enabled(lcore_id))
1118 lcore_ids[core_id++] = lcore_id;
1120 /* set the number of swithcing cores available */
1121 nb_switching_cores = rte_lcore_count()-1;
1123 /* Get the number of physical ports. */
1124 nb_ports = rte_eth_dev_count();
1125 if (nb_ports > RTE_MAX_ETHPORTS)
1126 nb_ports = RTE_MAX_ETHPORTS;
1129 * Update the global var NB_PORTS and global array PORTS
1130 * and get value of var VALID_NB_PORTS according to system ports number
1132 valid_nb_ports = check_ports_num(nb_ports);
1134 if ((valid_nb_ports == 0) || (valid_nb_ports > MAX_SUP_PORTS)) {
1135 rte_exit(EXIT_FAILURE, "Current enabled port number is %u,"
1136 "but only %u port can be enabled\n", nb_ports,
1139 /* Create the mbuf pool. */
1140 mbuf_pool = rte_mempool_create(
1144 MBUF_SIZE, MBUF_CACHE_SIZE,
1145 sizeof(struct rte_pktmbuf_pool_private),
1146 rte_pktmbuf_pool_init, NULL,
1147 rte_pktmbuf_init, NULL,
1148 rte_socket_id(), 0);
1149 if (mbuf_pool == NULL)
1150 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1152 for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
1153 vpool_array[queue_id].pool = mbuf_pool;
1155 /* Set log level. */
1156 rte_set_log_level(LOG_LEVEL);
1158 /* initialize all ports */
1159 for (portid = 0; portid < nb_ports; portid++) {
1160 /* skip ports that are not enabled */
1161 if ((enabled_port_mask & (1 << portid)) == 0) {
1162 RTE_LOG(INFO, VHOST_PORT,
1163 "Skipping disabled port %d\n", portid);
1166 if (overlay_options.port_configure(portid, mbuf_pool) != 0)
1167 rte_exit(EXIT_FAILURE,
1168 "Cannot initialize network ports\n");
1171 /* Initialise all linked lists. */
1172 if (init_data_ll() == -1)
1173 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1175 /* Initialize device stats */
1176 memset(&dev_statistics, 0, sizeof(dev_statistics));
1178 /* Enable stats if the user option is set. */
1180 pthread_create(&tid, NULL, (void *)print_stats, NULL);
1182 /* Launch all data cores. */
1183 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1184 rte_eal_remote_launch(switch_worker,
1185 mbuf_pool, lcore_id);
1187 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1189 /* Register CUSE device to handle IOCTLs. */
1190 ret = rte_vhost_driver_register((char *)&dev_basename);
1192 rte_exit(EXIT_FAILURE, "CUSE device setup failure.\n");
1194 rte_vhost_driver_callback_register(&virtio_net_device_ops);
1196 /* Start CUSE session. */
1197 rte_vhost_driver_session_start();