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_TSO_SEGSZ "tso-segsz"
118 #define CMD_LINE_OPT_FILTER_TYPE "filter-type"
119 #define CMD_LINE_OPT_RX_RETRY "rx-retry"
120 #define CMD_LINE_OPT_RX_RETRY_DELAY "rx-retry-delay"
121 #define CMD_LINE_OPT_RX_RETRY_NUM "rx-retry-num"
122 #define CMD_LINE_OPT_STATS "stats"
123 #define CMD_LINE_OPT_DEV_BASENAME "dev-basename"
125 /* mask of enabled ports */
126 static uint32_t enabled_port_mask;
128 /*Number of switching cores enabled*/
129 static uint32_t nb_switching_cores;
131 /* number of devices/queues to support*/
132 uint16_t nb_devices = 2;
134 /* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
135 #define MAX_RING_DESC 4096
138 struct rte_mempool *pool;
139 struct rte_ring *ring;
141 } vpool_array[MAX_QUEUES+MAX_QUEUES];
143 /* UDP tunneling port */
144 uint16_t udp_port = 4789;
146 /* enable/disable inner TX checksum */
147 uint8_t tx_checksum = 0;
149 /* TCP segment size */
150 uint16_t tso_segsz = 0;
152 /* RX filter type for tunneling packet */
153 uint8_t filter_idx = 1;
155 /* overlay packet operation */
156 struct ol_switch_ops overlay_options = {
157 .port_configure = vxlan_port_init,
158 .tunnel_setup = vxlan_link,
159 .tunnel_destroy = vxlan_unlink,
160 .tx_handle = vxlan_tx_pkts,
161 .rx_handle = vxlan_rx_pkts,
162 .param_handle = NULL,
166 uint32_t enable_stats = 0;
167 /* Enable retries on RX. */
168 static uint32_t enable_retry = 1;
169 /* Specify timeout (in useconds) between retries on RX. */
170 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
171 /* Specify the number of retries on RX. */
172 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
174 /* Character device basename. Can be set by user. */
175 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
177 static unsigned lcore_ids[RTE_MAX_LCORE];
178 uint8_t ports[RTE_MAX_ETHPORTS];
180 static unsigned nb_ports; /**< The number of ports specified in command line */
182 /* ethernet addresses of ports */
183 struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
185 /* heads for the main used and free linked lists for the data path. */
186 static struct virtio_net_data_ll *ll_root_used;
187 static struct virtio_net_data_ll *ll_root_free;
190 * Array of data core structures containing information on
191 * individual core linked lists.
193 static struct lcore_info lcore_info[RTE_MAX_LCORE];
195 /* Used for queueing bursts of TX packets. */
199 struct rte_mbuf *m_table[MAX_PKT_BURST];
202 /* TX queue for each data core. */
203 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
205 struct device_statistics dev_statistics[MAX_DEVICES];
208 * Set character device basename.
211 us_vhost_parse_basename(const char *q_arg)
213 /* parse number string */
214 if (strlen(q_arg) >= MAX_BASENAME_SZ)
217 snprintf((char *)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
223 * Parse the portmask provided at run time.
226 parse_portmask(const char *portmask)
231 /* parse hexadecimal string */
232 pm = strtoul(portmask, &end, 16);
233 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
243 * Parse num options at run time.
246 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
251 /* parse unsigned int string */
252 num = strtoul(q_arg, &end, 10);
253 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
256 if (num > max_valid_value)
266 tep_termination_usage(const char *prgname)
268 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
269 " --udp-port: UDP destination port for VXLAN packet\n"
270 " --nb-devices[1-64]: The number of virtIO device\n"
271 " --tx-checksum [0|1]: inner Tx checksum offload\n"
272 " --tso-segsz [0-N]: TCP segment size\n"
273 " --filter-type[1-3]: filter type for tunneling packet\n"
274 " 1: Inner MAC and tenent ID\n"
275 " 2: Inner MAC and VLAN, and tenent ID\n"
276 " 3: Outer MAC, Inner MAC and tenent ID\n"
277 " -p PORTMASK: Set mask for ports to be used by application\n"
278 " --rx-retry [0|1]: disable/enable(default) retries on rx."
279 " Enable retry if destintation queue is full\n"
280 " --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX."
281 " This makes effect only if retries on rx enabled\n"
282 " --rx-retry-num [0-N]: the number of retries on rx."
283 " This makes effect only if retries on rx enabled\n"
284 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
285 " --dev-basename: The basename to be used for the character device.\n",
290 * Parse the arguments given in the command line of the application.
293 tep_termination_parse_args(int argc, char **argv)
298 const char *prgname = argv[0];
299 static struct option long_option[] = {
300 {CMD_LINE_OPT_NB_DEVICES, required_argument, NULL, 0},
301 {CMD_LINE_OPT_UDP_PORT, required_argument, NULL, 0},
302 {CMD_LINE_OPT_TX_CHECKSUM, required_argument, NULL, 0},
303 {CMD_LINE_OPT_TSO_SEGSZ, required_argument, NULL, 0},
304 {CMD_LINE_OPT_FILTER_TYPE, required_argument, NULL, 0},
305 {CMD_LINE_OPT_RX_RETRY, required_argument, NULL, 0},
306 {CMD_LINE_OPT_RX_RETRY_DELAY, required_argument, NULL, 0},
307 {CMD_LINE_OPT_RX_RETRY_NUM, required_argument, NULL, 0},
308 {CMD_LINE_OPT_STATS, required_argument, NULL, 0},
309 {CMD_LINE_OPT_DEV_BASENAME, required_argument, NULL, 0},
313 /* Parse command line */
314 while ((opt = getopt_long(argc, argv, "p:",
315 long_option, &option_index)) != EOF) {
319 enabled_port_mask = parse_portmask(optarg);
320 if (enabled_port_mask == 0) {
321 RTE_LOG(INFO, VHOST_CONFIG,
322 "Invalid portmask\n");
323 tep_termination_usage(prgname);
328 if (!strncmp(long_option[option_index].name,
329 CMD_LINE_OPT_NB_DEVICES,
330 sizeof(CMD_LINE_OPT_NB_DEVICES))) {
331 ret = parse_num_opt(optarg, MAX_DEVICES);
333 RTE_LOG(INFO, VHOST_CONFIG,
334 "Invalid argument for nb-devices [0-%d]\n",
336 tep_termination_usage(prgname);
342 /* Enable/disable retries on RX. */
343 if (!strncmp(long_option[option_index].name,
344 CMD_LINE_OPT_RX_RETRY,
345 sizeof(CMD_LINE_OPT_RX_RETRY))) {
346 ret = parse_num_opt(optarg, 1);
348 RTE_LOG(INFO, VHOST_CONFIG,
349 "Invalid argument for rx-retry [0|1]\n");
350 tep_termination_usage(prgname);
356 if (!strncmp(long_option[option_index].name,
357 CMD_LINE_OPT_TSO_SEGSZ,
358 sizeof(CMD_LINE_OPT_TSO_SEGSZ))) {
359 ret = parse_num_opt(optarg, INT16_MAX);
361 RTE_LOG(INFO, VHOST_CONFIG,
362 "Invalid argument for TCP segment size [0-N]\n");
363 tep_termination_usage(prgname);
369 if (!strncmp(long_option[option_index].name,
370 CMD_LINE_OPT_UDP_PORT,
371 sizeof(CMD_LINE_OPT_UDP_PORT))) {
372 ret = parse_num_opt(optarg, INT16_MAX);
374 RTE_LOG(INFO, VHOST_CONFIG,
375 "Invalid argument for UDP port [0-N]\n");
376 tep_termination_usage(prgname);
382 /* Specify the retries delay time (in useconds) on RX.*/
383 if (!strncmp(long_option[option_index].name,
384 CMD_LINE_OPT_RX_RETRY_DELAY,
385 sizeof(CMD_LINE_OPT_RX_RETRY_DELAY))) {
386 ret = parse_num_opt(optarg, INT32_MAX);
388 RTE_LOG(INFO, VHOST_CONFIG,
389 "Invalid argument for rx-retry-delay [0-N]\n");
390 tep_termination_usage(prgname);
393 burst_rx_delay_time = ret;
396 /* Specify the retries number on RX. */
397 if (!strncmp(long_option[option_index].name,
398 CMD_LINE_OPT_RX_RETRY_NUM,
399 sizeof(CMD_LINE_OPT_RX_RETRY_NUM))) {
400 ret = parse_num_opt(optarg, INT32_MAX);
402 RTE_LOG(INFO, VHOST_CONFIG,
403 "Invalid argument for rx-retry-num [0-N]\n");
404 tep_termination_usage(prgname);
407 burst_rx_retry_num = ret;
410 if (!strncmp(long_option[option_index].name,
411 CMD_LINE_OPT_TX_CHECKSUM,
412 sizeof(CMD_LINE_OPT_TX_CHECKSUM))) {
413 ret = parse_num_opt(optarg, 1);
415 RTE_LOG(INFO, VHOST_CONFIG,
416 "Invalid argument for tx-checksum [0|1]\n");
417 tep_termination_usage(prgname);
423 if (!strncmp(long_option[option_index].name,
424 CMD_LINE_OPT_FILTER_TYPE,
425 sizeof(CMD_LINE_OPT_FILTER_TYPE))) {
426 ret = parse_num_opt(optarg, 3);
427 if ((ret == -1) || (ret == 0)) {
428 RTE_LOG(INFO, VHOST_CONFIG,
429 "Invalid argument for filter type [1-3]\n");
430 tep_termination_usage(prgname);
433 filter_idx = ret - 1;
436 /* Enable/disable stats. */
437 if (!strncmp(long_option[option_index].name,
439 sizeof(CMD_LINE_OPT_STATS))) {
440 ret = parse_num_opt(optarg, INT32_MAX);
442 RTE_LOG(INFO, VHOST_CONFIG,
443 "Invalid argument for stats [0..N]\n");
444 tep_termination_usage(prgname);
450 /* Set character device basename. */
451 if (!strncmp(long_option[option_index].name,
452 CMD_LINE_OPT_DEV_BASENAME,
453 sizeof(CMD_LINE_OPT_DEV_BASENAME))) {
454 if (us_vhost_parse_basename(optarg) == -1) {
455 RTE_LOG(INFO, VHOST_CONFIG,
456 "Invalid argument for character "
457 "device basename (Max %d characters)\n",
459 tep_termination_usage(prgname);
466 /* Invalid option - print options. */
468 tep_termination_usage(prgname);
473 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
474 if (enabled_port_mask & (1 << i))
475 ports[nb_ports++] = (uint8_t)i;
478 if ((nb_ports == 0) || (nb_ports > MAX_SUP_PORTS)) {
479 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
480 "but only %u port can be enabled\n", nb_ports,
489 * Update the global var NB_PORTS and array PORTS
490 * according to system ports number and return valid ports number
493 check_ports_num(unsigned max_nb_ports)
495 unsigned valid_nb_ports = nb_ports;
498 if (nb_ports > max_nb_ports) {
499 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) "
500 " exceeds total system port number(%u)\n",
501 nb_ports, max_nb_ports);
502 nb_ports = max_nb_ports;
505 for (portid = 0; portid < nb_ports; portid++) {
506 if (ports[portid] >= max_nb_ports) {
507 RTE_LOG(INFO, VHOST_PORT,
508 "\nSpecified port ID(%u) exceeds max "
509 " system port ID(%u)\n",
510 ports[portid], (max_nb_ports - 1));
511 ports[portid] = INVALID_PORT_ID;
515 return valid_nb_ports;
519 * This function routes the TX packet to the correct interface. This may be a local device
520 * or the physical port.
522 static inline void __attribute__((always_inline))
523 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
525 struct mbuf_table *tx_q;
526 struct rte_mbuf **m_table;
527 unsigned len, ret = 0;
528 const uint16_t lcore_id = rte_lcore_id();
529 struct virtio_net *dev = vdev->dev;
531 LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is external\n",
534 /* Add packet to the port tx queue */
535 tx_q = &lcore_tx_queue[lcore_id];
538 tx_q->m_table[len] = m;
541 dev_statistics[dev->device_fh].tx_total++;
542 dev_statistics[dev->device_fh].tx++;
545 if (unlikely(len == MAX_PKT_BURST)) {
546 m_table = (struct rte_mbuf **)tx_q->m_table;
547 ret = overlay_options.tx_handle(ports[0],
548 (uint16_t)tx_q->txq_id, m_table,
549 (uint16_t)tx_q->len);
551 /* Free any buffers not handled by TX and update
554 if (unlikely(ret < len)) {
556 rte_pktmbuf_free(m_table[ret]);
557 } while (++ret < len);
568 * This function is called by each data core. It handles all
569 * RX/TX registered with the core. For TX the specific lcore
570 * linked list is used. For RX, MAC addresses are compared
571 * with all devices in the main linked list.
574 switch_worker(__rte_unused void *arg)
576 struct rte_mempool *mbuf_pool = arg;
577 struct virtio_net *dev = NULL;
578 struct vhost_dev *vdev = NULL;
579 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
580 struct virtio_net_data_ll *dev_ll;
581 struct mbuf_table *tx_q;
582 volatile struct lcore_ll_info *lcore_ll;
583 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
584 / US_PER_S * BURST_TX_DRAIN_US;
585 uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
587 const uint16_t lcore_id = rte_lcore_id();
588 const uint16_t num_cores = (uint16_t)rte_lcore_count();
589 uint16_t rx_count = 0;
593 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
594 lcore_ll = lcore_info[lcore_id].lcore_ll;
597 tx_q = &lcore_tx_queue[lcore_id];
598 for (i = 0; i < num_cores; i++) {
599 if (lcore_ids[i] == lcore_id) {
606 cur_tsc = rte_rdtsc();
608 * TX burst queue drain
610 diff_tsc = cur_tsc - prev_tsc;
611 if (unlikely(diff_tsc > drain_tsc)) {
614 LOG_DEBUG(VHOST_DATA, "TX queue drained after "
615 "timeout with burst size %u\n",
617 ret = overlay_options.tx_handle(ports[0],
618 (uint16_t)tx_q->txq_id,
619 (struct rte_mbuf **)tx_q->m_table,
620 (uint16_t)tx_q->len);
621 if (unlikely(ret < tx_q->len)) {
623 rte_pktmbuf_free(tx_q->m_table[ret]);
624 } while (++ret < tx_q->len);
634 rte_prefetch0(lcore_ll->ll_root_used);
637 * Inform the configuration core that we have exited
638 * the linked list and that no devices are
639 * in use if requested.
641 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
642 lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
647 dev_ll = lcore_ll->ll_root_used;
649 while (dev_ll != NULL) {
653 if (unlikely(vdev->remove)) {
654 dev_ll = dev_ll->next;
655 overlay_options.tunnel_destroy(vdev);
656 vdev->ready = DEVICE_SAFE_REMOVE;
659 if (likely(vdev->ready == DEVICE_RX)) {
660 /* Handle guest RX */
661 rx_count = rte_eth_rx_burst(ports[0],
662 vdev->rx_q, pkts_burst, MAX_PKT_BURST);
666 * Retry is enabled and the queue is
667 * full then we wait and retry to
668 * avoid packet loss. Here MAX_PKT_BURST
669 * must be less than virtio queue size
671 if (enable_retry && unlikely(rx_count >
672 rte_vring_available_entries(dev, VIRTIO_RXQ))) {
673 for (retry = 0; retry < burst_rx_retry_num;
675 rte_delay_us(burst_rx_delay_time);
676 if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
681 ret_count = overlay_options.rx_handle(dev, pkts_burst, rx_count);
684 &dev_statistics[dev->device_fh].rx_total_atomic,
687 &dev_statistics[dev->device_fh].rx_atomic, ret_count);
689 while (likely(rx_count)) {
691 rte_pktmbuf_free(pkts_burst[rx_count]);
697 if (likely(!vdev->remove)) {
699 tx_count = rte_vhost_dequeue_burst(dev,
700 VIRTIO_TXQ, mbuf_pool,
701 pkts_burst, MAX_PKT_BURST);
702 /* If this is the first received packet we need to learn the MAC */
703 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
705 (overlay_options.tunnel_setup(vdev, pkts_burst[0]) == -1)) {
707 rte_pktmbuf_free(pkts_burst[--tx_count]);
711 virtio_tx_route(vdev, pkts_burst[--tx_count]);
714 /* move to the next device in the list */
715 dev_ll = dev_ll->next;
723 * Add an entry to a used linked list. A free entry must first be found
724 * in the free linked list using get_data_ll_free_entry();
727 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
728 struct virtio_net_data_ll *ll_dev)
730 struct virtio_net_data_ll *ll = *ll_root_addr;
732 /* Set next as NULL and use a compiler barrier to avoid reordering. */
734 rte_compiler_barrier();
736 /* If ll == NULL then this is the first device. */
738 /* Increment to the tail of the linked list. */
739 while (ll->next != NULL)
744 *ll_root_addr = ll_dev;
749 * Remove an entry from a used linked list. The entry must then be added to
750 * the free linked list using put_data_ll_free_entry().
753 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
754 struct virtio_net_data_ll *ll_dev,
755 struct virtio_net_data_ll *ll_dev_last)
757 struct virtio_net_data_ll *ll = *ll_root_addr;
759 if (unlikely((ll == NULL) || (ll_dev == NULL)))
763 *ll_root_addr = ll_dev->next;
765 if (likely(ll_dev_last != NULL))
766 ll_dev_last->next = ll_dev->next;
768 RTE_LOG(ERR, VHOST_CONFIG,
769 "Remove entry form ll failed.\n");
773 * Find and return an entry from the free linked list.
775 static struct virtio_net_data_ll *
776 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
778 struct virtio_net_data_ll *ll_free = *ll_root_addr;
779 struct virtio_net_data_ll *ll_dev;
785 *ll_root_addr = ll_free->next;
791 * Place an entry back on to the free linked list.
794 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
795 struct virtio_net_data_ll *ll_dev)
797 struct virtio_net_data_ll *ll_free = *ll_root_addr;
802 ll_dev->next = ll_free;
803 *ll_root_addr = ll_dev;
807 * Creates a linked list of a given size.
809 static struct virtio_net_data_ll *
810 alloc_data_ll(uint32_t size)
812 struct virtio_net_data_ll *ll_new;
815 /* Malloc and then chain the linked list. */
816 ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
817 if (ll_new == NULL) {
818 RTE_LOG(ERR, VHOST_CONFIG,
819 "Failed to allocate memory for ll_new.\n");
823 for (i = 0; i < size - 1; i++) {
824 ll_new[i].vdev = NULL;
825 ll_new[i].next = &ll_new[i+1];
827 ll_new[i].next = NULL;
833 * Create the main linked list along with each individual cores
834 * linked list. A used and a free list are created to manage entries.
841 RTE_LCORE_FOREACH_SLAVE(lcore) {
842 lcore_info[lcore].lcore_ll =
843 malloc(sizeof(struct lcore_ll_info));
844 if (lcore_info[lcore].lcore_ll == NULL) {
845 RTE_LOG(ERR, VHOST_CONFIG,
846 "Failed to allocate memory for lcore_ll.\n");
850 lcore_info[lcore].lcore_ll->device_num = 0;
851 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
852 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
853 if (nb_devices % nb_switching_cores)
854 lcore_info[lcore].lcore_ll->ll_root_free =
855 alloc_data_ll((nb_devices / nb_switching_cores)
858 lcore_info[lcore].lcore_ll->ll_root_free =
859 alloc_data_ll(nb_devices / nb_switching_cores);
862 /* Allocate devices up to a maximum of MAX_DEVICES. */
863 ll_root_free = alloc_data_ll(MIN((nb_devices), MAX_DEVICES));
869 * Remove a device from the specific data core linked list and
870 * from the main linked list. Synchonization occurs through the use
871 * of the lcore dev_removal_flag. Device is made volatile here
872 * to avoid re-ordering of dev->remove=1 which can cause an infinite
873 * loop in the rte_pause loop.
876 destroy_device(volatile struct virtio_net *dev)
878 struct virtio_net_data_ll *ll_lcore_dev_cur;
879 struct virtio_net_data_ll *ll_main_dev_cur;
880 struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
881 struct virtio_net_data_ll *ll_main_dev_last = NULL;
882 struct vhost_dev *vdev;
885 dev->flags &= ~VIRTIO_DEV_RUNNING;
887 vdev = (struct vhost_dev *)dev->priv;
889 /* set the remove flag. */
891 while (vdev->ready != DEVICE_SAFE_REMOVE)
894 /* Search for entry to be removed from lcore ll */
895 ll_lcore_dev_cur = lcore_info[vdev->coreid].lcore_ll->ll_root_used;
896 while (ll_lcore_dev_cur != NULL) {
897 if (ll_lcore_dev_cur->vdev == vdev) {
900 ll_lcore_dev_last = ll_lcore_dev_cur;
901 ll_lcore_dev_cur = ll_lcore_dev_cur->next;
905 if (ll_lcore_dev_cur == NULL) {
906 RTE_LOG(ERR, VHOST_CONFIG,
907 "(%"PRIu64") Failed to find the dev to be destroy.\n",
912 /* Search for entry to be removed from main ll */
913 ll_main_dev_cur = ll_root_used;
914 ll_main_dev_last = NULL;
915 while (ll_main_dev_cur != NULL) {
916 if (ll_main_dev_cur->vdev == vdev) {
919 ll_main_dev_last = ll_main_dev_cur;
920 ll_main_dev_cur = ll_main_dev_cur->next;
924 /* Remove entries from the lcore and main ll. */
925 rm_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
926 ll_lcore_dev_cur, ll_lcore_dev_last);
927 rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
929 /* Set the dev_removal_flag on each lcore. */
930 RTE_LCORE_FOREACH_SLAVE(lcore) {
931 lcore_info[lcore].lcore_ll->dev_removal_flag =
936 * Once each core has set the dev_removal_flag to
937 * ACK_DEV_REMOVAL we can be sure that they can no longer access
938 * the device removed from the linked lists and that the devices
939 * are no longer in use.
941 RTE_LCORE_FOREACH_SLAVE(lcore) {
942 while (lcore_info[lcore].lcore_ll->dev_removal_flag
947 /* Add the entries back to the lcore and main free ll.*/
948 put_data_ll_free_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_free,
950 put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
952 /* Decrement number of device on the lcore. */
953 lcore_info[vdev->coreid].lcore_ll->device_num--;
955 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been removed "
956 "from data core\n", dev->device_fh);
963 * A new device is added to a data core. First the device is added
964 * to the main linked list and the allocated to a specific data core.
967 new_device(struct virtio_net *dev)
969 struct virtio_net_data_ll *ll_dev;
970 int lcore, core_add = 0;
971 uint32_t device_num_min = nb_devices;
972 struct vhost_dev *vdev;
974 vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
976 RTE_LOG(INFO, VHOST_DATA,
977 "(%"PRIu64") Couldn't allocate memory for vhost dev\n",
983 /* Add device to main ll */
984 ll_dev = get_data_ll_free_entry(&ll_root_free);
985 if (ll_dev == NULL) {
986 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in"
987 " linked list Device limit of %d devices per core"
988 " has been reached\n", dev->device_fh, nb_devices);
989 if (vdev->regions_hpa)
990 rte_free(vdev->regions_hpa);
995 add_data_ll_entry(&ll_root_used, ll_dev);
996 vdev->rx_q = dev->device_fh;
998 /* reset ready flag */
999 vdev->ready = DEVICE_MAC_LEARNING;
1002 /* Find a suitable lcore to add the device. */
1003 RTE_LCORE_FOREACH_SLAVE(lcore) {
1004 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
1005 device_num_min = lcore_info[lcore].lcore_ll->device_num;
1009 /* Add device to lcore ll */
1010 ll_dev = get_data_ll_free_entry(&lcore_info[core_add].lcore_ll->ll_root_free);
1011 if (ll_dev == NULL) {
1012 RTE_LOG(INFO, VHOST_DATA,
1013 "(%"PRIu64") Failed to add device to data core\n",
1015 vdev->ready = DEVICE_SAFE_REMOVE;
1016 destroy_device(dev);
1017 rte_free(vdev->regions_hpa);
1021 ll_dev->vdev = vdev;
1022 vdev->coreid = core_add;
1024 add_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
1027 /* Initialize device stats */
1028 memset(&dev_statistics[dev->device_fh], 0,
1029 sizeof(struct device_statistics));
1031 /* Disable notifications. */
1032 rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
1033 rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
1034 lcore_info[vdev->coreid].lcore_ll->device_num++;
1035 dev->flags |= VIRTIO_DEV_RUNNING;
1037 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n",
1038 dev->device_fh, vdev->coreid);
1044 * These callback allow devices to be added to the data core when configuration
1045 * has been fully complete.
1047 static const struct virtio_net_device_ops virtio_net_device_ops = {
1048 .new_device = new_device,
1049 .destroy_device = destroy_device,
1053 * This is a thread will wake up after a period to print stats if the user has
1059 struct virtio_net_data_ll *dev_ll;
1060 uint64_t tx_dropped, rx_dropped;
1061 uint64_t tx, tx_total, rx, rx_total;
1063 const char clr[] = { 27, '[', '2', 'J', '\0' };
1064 const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
1067 sleep(enable_stats);
1069 /* Clear screen and move to top left */
1070 printf("%s%s", clr, top_left);
1072 printf("\nDevice statistics ================================");
1074 dev_ll = ll_root_used;
1075 while (dev_ll != NULL) {
1076 device_fh = (uint32_t)dev_ll->vdev->dev->device_fh;
1077 tx_total = dev_statistics[device_fh].tx_total;
1078 tx = dev_statistics[device_fh].tx;
1079 tx_dropped = tx_total - tx;
1081 rx_total = rte_atomic64_read(
1082 &dev_statistics[device_fh].rx_total_atomic);
1083 rx = rte_atomic64_read(
1084 &dev_statistics[device_fh].rx_atomic);
1085 rx_dropped = rx_total - rx;
1087 printf("\nStatistics for device %"PRIu32" ----------"
1088 "\nTX total: %"PRIu64""
1089 "\nTX dropped: %"PRIu64""
1090 "\nTX successful: %"PRIu64""
1091 "\nRX total: %"PRIu64""
1092 "\nRX dropped: %"PRIu64""
1093 "\nRX successful: %"PRIu64"",
1102 dev_ll = dev_ll->next;
1104 printf("\n================================================\n");
1109 * Main function, does initialisation and calls the per-lcore functions. The CUSE
1110 * device is also registered here to handle the IOCTLs.
1113 main(int argc, char *argv[])
1115 struct rte_mempool *mbuf_pool = NULL;
1116 unsigned lcore_id, core_id = 0;
1117 unsigned nb_ports, valid_nb_ports;
1121 static pthread_t tid;
1124 ret = rte_eal_init(argc, argv);
1126 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1130 /* parse app arguments */
1131 ret = tep_termination_parse_args(argc, argv);
1133 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1135 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
1136 if (rte_lcore_is_enabled(lcore_id))
1137 lcore_ids[core_id++] = lcore_id;
1139 /* set the number of swithcing cores available */
1140 nb_switching_cores = rte_lcore_count()-1;
1142 /* Get the number of physical ports. */
1143 nb_ports = rte_eth_dev_count();
1144 if (nb_ports > RTE_MAX_ETHPORTS)
1145 nb_ports = RTE_MAX_ETHPORTS;
1148 * Update the global var NB_PORTS and global array PORTS
1149 * and get value of var VALID_NB_PORTS according to system ports number
1151 valid_nb_ports = check_ports_num(nb_ports);
1153 if ((valid_nb_ports == 0) || (valid_nb_ports > MAX_SUP_PORTS)) {
1154 rte_exit(EXIT_FAILURE, "Current enabled port number is %u,"
1155 "but only %u port can be enabled\n", nb_ports,
1158 /* Create the mbuf pool. */
1159 mbuf_pool = rte_mempool_create(
1163 MBUF_SIZE, MBUF_CACHE_SIZE,
1164 sizeof(struct rte_pktmbuf_pool_private),
1165 rte_pktmbuf_pool_init, NULL,
1166 rte_pktmbuf_init, NULL,
1167 rte_socket_id(), 0);
1168 if (mbuf_pool == NULL)
1169 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1171 for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
1172 vpool_array[queue_id].pool = mbuf_pool;
1174 /* Set log level. */
1175 rte_set_log_level(LOG_LEVEL);
1177 /* initialize all ports */
1178 for (portid = 0; portid < nb_ports; portid++) {
1179 /* skip ports that are not enabled */
1180 if ((enabled_port_mask & (1 << portid)) == 0) {
1181 RTE_LOG(INFO, VHOST_PORT,
1182 "Skipping disabled port %d\n", portid);
1185 if (overlay_options.port_configure(portid, mbuf_pool) != 0)
1186 rte_exit(EXIT_FAILURE,
1187 "Cannot initialize network ports\n");
1190 /* Initialise all linked lists. */
1191 if (init_data_ll() == -1)
1192 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1194 /* Initialize device stats */
1195 memset(&dev_statistics, 0, sizeof(dev_statistics));
1197 /* Enable stats if the user option is set. */
1199 pthread_create(&tid, NULL, (void *)print_stats, NULL);
1201 /* Launch all data cores. */
1202 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1203 rte_eal_remote_launch(switch_worker,
1204 mbuf_pool, lcore_id);
1206 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1208 /* Register CUSE device to handle IOCTLs. */
1209 ret = rte_vhost_driver_register((char *)&dev_basename);
1211 rte_exit(EXIT_FAILURE, "CUSE device setup failure.\n");
1213 rte_vhost_driver_callback_register(&virtio_net_device_ops);
1215 /* Start CUSE session. */
1216 rte_vhost_driver_session_start();