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_ENCAP "encap"
120 #define CMD_LINE_OPT_DECAP "decap"
121 #define CMD_LINE_OPT_RX_RETRY "rx-retry"
122 #define CMD_LINE_OPT_RX_RETRY_DELAY "rx-retry-delay"
123 #define CMD_LINE_OPT_RX_RETRY_NUM "rx-retry-num"
124 #define CMD_LINE_OPT_STATS "stats"
125 #define CMD_LINE_OPT_DEV_BASENAME "dev-basename"
127 /* mask of enabled ports */
128 static uint32_t enabled_port_mask;
130 /*Number of switching cores enabled*/
131 static uint32_t nb_switching_cores;
133 /* number of devices/queues to support*/
134 uint16_t nb_devices = 2;
136 /* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
137 #define MAX_RING_DESC 4096
140 struct rte_mempool *pool;
141 struct rte_ring *ring;
143 } vpool_array[MAX_QUEUES+MAX_QUEUES];
145 /* UDP tunneling port */
146 uint16_t udp_port = 4789;
148 /* enable/disable inner TX checksum */
149 uint8_t tx_checksum = 0;
151 /* TCP segment size */
152 uint16_t tso_segsz = 0;
154 /* enable/disable decapsulation */
155 uint8_t rx_decap = 1;
157 /* enable/disable encapsulation */
158 uint8_t tx_encap = 1;
160 /* RX filter type for tunneling packet */
161 uint8_t filter_idx = 1;
163 /* overlay packet operation */
164 struct ol_switch_ops overlay_options = {
165 .port_configure = vxlan_port_init,
166 .tunnel_setup = vxlan_link,
167 .tunnel_destroy = vxlan_unlink,
168 .tx_handle = vxlan_tx_pkts,
169 .rx_handle = vxlan_rx_pkts,
170 .param_handle = NULL,
174 uint32_t enable_stats = 0;
175 /* Enable retries on RX. */
176 static uint32_t enable_retry = 1;
177 /* Specify timeout (in useconds) between retries on RX. */
178 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
179 /* Specify the number of retries on RX. */
180 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
182 /* Character device basename. Can be set by user. */
183 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
185 static unsigned lcore_ids[RTE_MAX_LCORE];
186 uint8_t ports[RTE_MAX_ETHPORTS];
188 static unsigned nb_ports; /**< The number of ports specified in command line */
190 /* ethernet addresses of ports */
191 struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
193 /* heads for the main used and free linked lists for the data path. */
194 static struct virtio_net_data_ll *ll_root_used;
195 static struct virtio_net_data_ll *ll_root_free;
198 * Array of data core structures containing information on
199 * individual core linked lists.
201 static struct lcore_info lcore_info[RTE_MAX_LCORE];
203 /* Used for queueing bursts of TX packets. */
207 struct rte_mbuf *m_table[MAX_PKT_BURST];
210 /* TX queue for each data core. */
211 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
213 struct device_statistics dev_statistics[MAX_DEVICES];
216 * Set character device basename.
219 us_vhost_parse_basename(const char *q_arg)
221 /* parse number string */
222 if (strlen(q_arg) >= MAX_BASENAME_SZ)
225 snprintf((char *)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
231 * Parse the portmask provided at run time.
234 parse_portmask(const char *portmask)
239 /* parse hexadecimal string */
240 pm = strtoul(portmask, &end, 16);
241 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
251 * Parse num options at run time.
254 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
259 /* parse unsigned int string */
260 num = strtoul(q_arg, &end, 10);
261 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
264 if (num > max_valid_value)
274 tep_termination_usage(const char *prgname)
276 RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
277 " --udp-port: UDP destination port for VXLAN packet\n"
278 " --nb-devices[1-64]: The number of virtIO device\n"
279 " --tx-checksum [0|1]: inner Tx checksum offload\n"
280 " --tso-segsz [0-N]: TCP segment size\n"
281 " --decap [0|1]: tunneling packet decapsulation\n"
282 " --encap [0|1]: tunneling packet encapsulation\n"
283 " --filter-type[1-3]: filter type for tunneling packet\n"
284 " 1: Inner MAC and tenent ID\n"
285 " 2: Inner MAC and VLAN, and tenent ID\n"
286 " 3: Outer MAC, Inner MAC and tenent ID\n"
287 " -p PORTMASK: Set mask for ports to be used by application\n"
288 " --rx-retry [0|1]: disable/enable(default) retries on rx."
289 " Enable retry if destintation queue is full\n"
290 " --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX."
291 " This makes effect only if retries on rx enabled\n"
292 " --rx-retry-num [0-N]: the number of retries on rx."
293 " This makes effect only if retries on rx enabled\n"
294 " --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
295 " --dev-basename: The basename to be used for the character device.\n",
300 * Parse the arguments given in the command line of the application.
303 tep_termination_parse_args(int argc, char **argv)
308 const char *prgname = argv[0];
309 static struct option long_option[] = {
310 {CMD_LINE_OPT_NB_DEVICES, required_argument, NULL, 0},
311 {CMD_LINE_OPT_UDP_PORT, required_argument, NULL, 0},
312 {CMD_LINE_OPT_TX_CHECKSUM, required_argument, NULL, 0},
313 {CMD_LINE_OPT_TSO_SEGSZ, required_argument, NULL, 0},
314 {CMD_LINE_OPT_DECAP, required_argument, NULL, 0},
315 {CMD_LINE_OPT_ENCAP, required_argument, NULL, 0},
316 {CMD_LINE_OPT_FILTER_TYPE, required_argument, NULL, 0},
317 {CMD_LINE_OPT_RX_RETRY, required_argument, NULL, 0},
318 {CMD_LINE_OPT_RX_RETRY_DELAY, required_argument, NULL, 0},
319 {CMD_LINE_OPT_RX_RETRY_NUM, required_argument, NULL, 0},
320 {CMD_LINE_OPT_STATS, required_argument, NULL, 0},
321 {CMD_LINE_OPT_DEV_BASENAME, required_argument, NULL, 0},
325 /* Parse command line */
326 while ((opt = getopt_long(argc, argv, "p:",
327 long_option, &option_index)) != EOF) {
331 enabled_port_mask = parse_portmask(optarg);
332 if (enabled_port_mask == 0) {
333 RTE_LOG(INFO, VHOST_CONFIG,
334 "Invalid portmask\n");
335 tep_termination_usage(prgname);
340 if (!strncmp(long_option[option_index].name,
341 CMD_LINE_OPT_NB_DEVICES,
342 sizeof(CMD_LINE_OPT_NB_DEVICES))) {
343 ret = parse_num_opt(optarg, MAX_DEVICES);
345 RTE_LOG(INFO, VHOST_CONFIG,
346 "Invalid argument for nb-devices [0-%d]\n",
348 tep_termination_usage(prgname);
354 /* Enable/disable retries on RX. */
355 if (!strncmp(long_option[option_index].name,
356 CMD_LINE_OPT_RX_RETRY,
357 sizeof(CMD_LINE_OPT_RX_RETRY))) {
358 ret = parse_num_opt(optarg, 1);
360 RTE_LOG(INFO, VHOST_CONFIG,
361 "Invalid argument for rx-retry [0|1]\n");
362 tep_termination_usage(prgname);
368 if (!strncmp(long_option[option_index].name,
369 CMD_LINE_OPT_TSO_SEGSZ,
370 sizeof(CMD_LINE_OPT_TSO_SEGSZ))) {
371 ret = parse_num_opt(optarg, INT16_MAX);
373 RTE_LOG(INFO, VHOST_CONFIG,
374 "Invalid argument for TCP segment size [0-N]\n");
375 tep_termination_usage(prgname);
381 if (!strncmp(long_option[option_index].name,
382 CMD_LINE_OPT_UDP_PORT,
383 sizeof(CMD_LINE_OPT_UDP_PORT))) {
384 ret = parse_num_opt(optarg, INT16_MAX);
386 RTE_LOG(INFO, VHOST_CONFIG,
387 "Invalid argument for UDP port [0-N]\n");
388 tep_termination_usage(prgname);
394 /* Specify the retries delay time (in useconds) on RX.*/
395 if (!strncmp(long_option[option_index].name,
396 CMD_LINE_OPT_RX_RETRY_DELAY,
397 sizeof(CMD_LINE_OPT_RX_RETRY_DELAY))) {
398 ret = parse_num_opt(optarg, INT32_MAX);
400 RTE_LOG(INFO, VHOST_CONFIG,
401 "Invalid argument for rx-retry-delay [0-N]\n");
402 tep_termination_usage(prgname);
405 burst_rx_delay_time = ret;
408 /* Specify the retries number on RX. */
409 if (!strncmp(long_option[option_index].name,
410 CMD_LINE_OPT_RX_RETRY_NUM,
411 sizeof(CMD_LINE_OPT_RX_RETRY_NUM))) {
412 ret = parse_num_opt(optarg, INT32_MAX);
414 RTE_LOG(INFO, VHOST_CONFIG,
415 "Invalid argument for rx-retry-num [0-N]\n");
416 tep_termination_usage(prgname);
419 burst_rx_retry_num = ret;
422 if (!strncmp(long_option[option_index].name,
423 CMD_LINE_OPT_TX_CHECKSUM,
424 sizeof(CMD_LINE_OPT_TX_CHECKSUM))) {
425 ret = parse_num_opt(optarg, 1);
427 RTE_LOG(INFO, VHOST_CONFIG,
428 "Invalid argument for tx-checksum [0|1]\n");
429 tep_termination_usage(prgname);
435 if (!strncmp(long_option[option_index].name,
436 CMD_LINE_OPT_FILTER_TYPE,
437 sizeof(CMD_LINE_OPT_FILTER_TYPE))) {
438 ret = parse_num_opt(optarg, 3);
439 if ((ret == -1) || (ret == 0)) {
440 RTE_LOG(INFO, VHOST_CONFIG,
441 "Invalid argument for filter type [1-3]\n");
442 tep_termination_usage(prgname);
445 filter_idx = ret - 1;
448 /* Enable/disable encapsulation on RX. */
449 if (!strncmp(long_option[option_index].name,
451 sizeof(CMD_LINE_OPT_DECAP))) {
452 ret = parse_num_opt(optarg, 1);
454 RTE_LOG(INFO, VHOST_CONFIG,
455 "Invalid argument for decap [0|1]\n");
456 tep_termination_usage(prgname);
462 /* Enable/disable encapsulation on TX. */
463 if (!strncmp(long_option[option_index].name,
465 sizeof(CMD_LINE_OPT_ENCAP))) {
466 ret = parse_num_opt(optarg, 1);
468 RTE_LOG(INFO, VHOST_CONFIG,
469 "Invalid argument for encap [0|1]\n");
470 tep_termination_usage(prgname);
476 /* Enable/disable stats. */
477 if (!strncmp(long_option[option_index].name,
479 sizeof(CMD_LINE_OPT_STATS))) {
480 ret = parse_num_opt(optarg, INT32_MAX);
482 RTE_LOG(INFO, VHOST_CONFIG,
483 "Invalid argument for stats [0..N]\n");
484 tep_termination_usage(prgname);
490 /* Set character device basename. */
491 if (!strncmp(long_option[option_index].name,
492 CMD_LINE_OPT_DEV_BASENAME,
493 sizeof(CMD_LINE_OPT_DEV_BASENAME))) {
494 if (us_vhost_parse_basename(optarg) == -1) {
495 RTE_LOG(INFO, VHOST_CONFIG,
496 "Invalid argument for character "
497 "device basename (Max %d characters)\n",
499 tep_termination_usage(prgname);
506 /* Invalid option - print options. */
508 tep_termination_usage(prgname);
513 for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
514 if (enabled_port_mask & (1 << i))
515 ports[nb_ports++] = (uint8_t)i;
518 if ((nb_ports == 0) || (nb_ports > MAX_SUP_PORTS)) {
519 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
520 "but only %u port can be enabled\n", nb_ports,
529 * Update the global var NB_PORTS and array PORTS
530 * according to system ports number and return valid ports number
533 check_ports_num(unsigned max_nb_ports)
535 unsigned valid_nb_ports = nb_ports;
538 if (nb_ports > max_nb_ports) {
539 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) "
540 " exceeds total system port number(%u)\n",
541 nb_ports, max_nb_ports);
542 nb_ports = max_nb_ports;
545 for (portid = 0; portid < nb_ports; portid++) {
546 if (ports[portid] >= max_nb_ports) {
547 RTE_LOG(INFO, VHOST_PORT,
548 "\nSpecified port ID(%u) exceeds max "
549 " system port ID(%u)\n",
550 ports[portid], (max_nb_ports - 1));
551 ports[portid] = INVALID_PORT_ID;
555 return valid_nb_ports;
559 * This function routes the TX packet to the correct interface. This may be a local device
560 * or the physical port.
562 static inline void __attribute__((always_inline))
563 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
565 struct mbuf_table *tx_q;
566 struct rte_mbuf **m_table;
567 unsigned len, ret = 0;
568 const uint16_t lcore_id = rte_lcore_id();
569 struct virtio_net *dev = vdev->dev;
571 LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is external\n",
574 /* Add packet to the port tx queue */
575 tx_q = &lcore_tx_queue[lcore_id];
578 tx_q->m_table[len] = m;
581 dev_statistics[dev->device_fh].tx_total++;
582 dev_statistics[dev->device_fh].tx++;
585 if (unlikely(len == MAX_PKT_BURST)) {
586 m_table = (struct rte_mbuf **)tx_q->m_table;
587 ret = overlay_options.tx_handle(ports[0],
588 (uint16_t)tx_q->txq_id, m_table,
589 (uint16_t)tx_q->len);
591 /* Free any buffers not handled by TX and update
594 if (unlikely(ret < len)) {
596 rte_pktmbuf_free(m_table[ret]);
597 } while (++ret < len);
608 * This function is called by each data core. It handles all
609 * RX/TX registered with the core. For TX the specific lcore
610 * linked list is used. For RX, MAC addresses are compared
611 * with all devices in the main linked list.
614 switch_worker(__rte_unused void *arg)
616 struct rte_mempool *mbuf_pool = arg;
617 struct virtio_net *dev = NULL;
618 struct vhost_dev *vdev = NULL;
619 struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
620 struct virtio_net_data_ll *dev_ll;
621 struct mbuf_table *tx_q;
622 volatile struct lcore_ll_info *lcore_ll;
623 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
624 / US_PER_S * BURST_TX_DRAIN_US;
625 uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
627 const uint16_t lcore_id = rte_lcore_id();
628 const uint16_t num_cores = (uint16_t)rte_lcore_count();
629 uint16_t rx_count = 0;
633 RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
634 lcore_ll = lcore_info[lcore_id].lcore_ll;
637 tx_q = &lcore_tx_queue[lcore_id];
638 for (i = 0; i < num_cores; i++) {
639 if (lcore_ids[i] == lcore_id) {
646 cur_tsc = rte_rdtsc();
648 * TX burst queue drain
650 diff_tsc = cur_tsc - prev_tsc;
651 if (unlikely(diff_tsc > drain_tsc)) {
654 LOG_DEBUG(VHOST_DATA, "TX queue drained after "
655 "timeout with burst size %u\n",
657 ret = overlay_options.tx_handle(ports[0],
658 (uint16_t)tx_q->txq_id,
659 (struct rte_mbuf **)tx_q->m_table,
660 (uint16_t)tx_q->len);
661 if (unlikely(ret < tx_q->len)) {
663 rte_pktmbuf_free(tx_q->m_table[ret]);
664 } while (++ret < tx_q->len);
674 rte_prefetch0(lcore_ll->ll_root_used);
677 * Inform the configuration core that we have exited
678 * the linked list and that no devices are
679 * in use if requested.
681 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
682 lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
687 dev_ll = lcore_ll->ll_root_used;
689 while (dev_ll != NULL) {
693 if (unlikely(vdev->remove)) {
694 dev_ll = dev_ll->next;
695 overlay_options.tunnel_destroy(vdev);
696 vdev->ready = DEVICE_SAFE_REMOVE;
699 if (likely(vdev->ready == DEVICE_RX)) {
700 /* Handle guest RX */
701 rx_count = rte_eth_rx_burst(ports[0],
702 vdev->rx_q, pkts_burst, MAX_PKT_BURST);
706 * Retry is enabled and the queue is
707 * full then we wait and retry to
708 * avoid packet loss. Here MAX_PKT_BURST
709 * must be less than virtio queue size
711 if (enable_retry && unlikely(rx_count >
712 rte_vring_available_entries(dev, VIRTIO_RXQ))) {
713 for (retry = 0; retry < burst_rx_retry_num;
715 rte_delay_us(burst_rx_delay_time);
716 if (rx_count <= rte_vring_available_entries(dev, VIRTIO_RXQ))
721 ret_count = overlay_options.rx_handle(dev, pkts_burst, rx_count);
724 &dev_statistics[dev->device_fh].rx_total_atomic,
727 &dev_statistics[dev->device_fh].rx_atomic, ret_count);
729 while (likely(rx_count)) {
731 rte_pktmbuf_free(pkts_burst[rx_count]);
737 if (likely(!vdev->remove)) {
739 tx_count = rte_vhost_dequeue_burst(dev,
740 VIRTIO_TXQ, mbuf_pool,
741 pkts_burst, MAX_PKT_BURST);
742 /* If this is the first received packet we need to learn the MAC */
743 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
745 (overlay_options.tunnel_setup(vdev, pkts_burst[0]) == -1)) {
747 rte_pktmbuf_free(pkts_burst[--tx_count]);
751 virtio_tx_route(vdev, pkts_burst[--tx_count]);
754 /* move to the next device in the list */
755 dev_ll = dev_ll->next;
763 * Add an entry to a used linked list. A free entry must first be found
764 * in the free linked list using get_data_ll_free_entry();
767 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
768 struct virtio_net_data_ll *ll_dev)
770 struct virtio_net_data_ll *ll = *ll_root_addr;
772 /* Set next as NULL and use a compiler barrier to avoid reordering. */
774 rte_compiler_barrier();
776 /* If ll == NULL then this is the first device. */
778 /* Increment to the tail of the linked list. */
779 while (ll->next != NULL)
784 *ll_root_addr = ll_dev;
789 * Remove an entry from a used linked list. The entry must then be added to
790 * the free linked list using put_data_ll_free_entry().
793 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
794 struct virtio_net_data_ll *ll_dev,
795 struct virtio_net_data_ll *ll_dev_last)
797 struct virtio_net_data_ll *ll = *ll_root_addr;
799 if (unlikely((ll == NULL) || (ll_dev == NULL)))
803 *ll_root_addr = ll_dev->next;
805 if (likely(ll_dev_last != NULL))
806 ll_dev_last->next = ll_dev->next;
808 RTE_LOG(ERR, VHOST_CONFIG,
809 "Remove entry form ll failed.\n");
813 * Find and return an entry from the free linked list.
815 static struct virtio_net_data_ll *
816 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
818 struct virtio_net_data_ll *ll_free = *ll_root_addr;
819 struct virtio_net_data_ll *ll_dev;
825 *ll_root_addr = ll_free->next;
831 * Place an entry back on to the free linked list.
834 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
835 struct virtio_net_data_ll *ll_dev)
837 struct virtio_net_data_ll *ll_free = *ll_root_addr;
842 ll_dev->next = ll_free;
843 *ll_root_addr = ll_dev;
847 * Creates a linked list of a given size.
849 static struct virtio_net_data_ll *
850 alloc_data_ll(uint32_t size)
852 struct virtio_net_data_ll *ll_new;
855 /* Malloc and then chain the linked list. */
856 ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
857 if (ll_new == NULL) {
858 RTE_LOG(ERR, VHOST_CONFIG,
859 "Failed to allocate memory for ll_new.\n");
863 for (i = 0; i < size - 1; i++) {
864 ll_new[i].vdev = NULL;
865 ll_new[i].next = &ll_new[i+1];
867 ll_new[i].next = NULL;
873 * Create the main linked list along with each individual cores
874 * linked list. A used and a free list are created to manage entries.
881 RTE_LCORE_FOREACH_SLAVE(lcore) {
882 lcore_info[lcore].lcore_ll =
883 malloc(sizeof(struct lcore_ll_info));
884 if (lcore_info[lcore].lcore_ll == NULL) {
885 RTE_LOG(ERR, VHOST_CONFIG,
886 "Failed to allocate memory for lcore_ll.\n");
890 lcore_info[lcore].lcore_ll->device_num = 0;
891 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
892 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
893 if (nb_devices % nb_switching_cores)
894 lcore_info[lcore].lcore_ll->ll_root_free =
895 alloc_data_ll((nb_devices / nb_switching_cores)
898 lcore_info[lcore].lcore_ll->ll_root_free =
899 alloc_data_ll(nb_devices / nb_switching_cores);
902 /* Allocate devices up to a maximum of MAX_DEVICES. */
903 ll_root_free = alloc_data_ll(MIN((nb_devices), MAX_DEVICES));
909 * Remove a device from the specific data core linked list and
910 * from the main linked list. Synchonization occurs through the use
911 * of the lcore dev_removal_flag. Device is made volatile here
912 * to avoid re-ordering of dev->remove=1 which can cause an infinite
913 * loop in the rte_pause loop.
916 destroy_device(volatile struct virtio_net *dev)
918 struct virtio_net_data_ll *ll_lcore_dev_cur;
919 struct virtio_net_data_ll *ll_main_dev_cur;
920 struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
921 struct virtio_net_data_ll *ll_main_dev_last = NULL;
922 struct vhost_dev *vdev;
925 dev->flags &= ~VIRTIO_DEV_RUNNING;
927 vdev = (struct vhost_dev *)dev->priv;
929 /* set the remove flag. */
931 while (vdev->ready != DEVICE_SAFE_REMOVE)
934 /* Search for entry to be removed from lcore ll */
935 ll_lcore_dev_cur = lcore_info[vdev->coreid].lcore_ll->ll_root_used;
936 while (ll_lcore_dev_cur != NULL) {
937 if (ll_lcore_dev_cur->vdev == vdev) {
940 ll_lcore_dev_last = ll_lcore_dev_cur;
941 ll_lcore_dev_cur = ll_lcore_dev_cur->next;
945 if (ll_lcore_dev_cur == NULL) {
946 RTE_LOG(ERR, VHOST_CONFIG,
947 "(%"PRIu64") Failed to find the dev to be destroy.\n",
952 /* Search for entry to be removed from main ll */
953 ll_main_dev_cur = ll_root_used;
954 ll_main_dev_last = NULL;
955 while (ll_main_dev_cur != NULL) {
956 if (ll_main_dev_cur->vdev == vdev) {
959 ll_main_dev_last = ll_main_dev_cur;
960 ll_main_dev_cur = ll_main_dev_cur->next;
964 /* Remove entries from the lcore and main ll. */
965 rm_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
966 ll_lcore_dev_cur, ll_lcore_dev_last);
967 rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
969 /* Set the dev_removal_flag on each lcore. */
970 RTE_LCORE_FOREACH_SLAVE(lcore) {
971 lcore_info[lcore].lcore_ll->dev_removal_flag =
976 * Once each core has set the dev_removal_flag to
977 * ACK_DEV_REMOVAL we can be sure that they can no longer access
978 * the device removed from the linked lists and that the devices
979 * are no longer in use.
981 RTE_LCORE_FOREACH_SLAVE(lcore) {
982 while (lcore_info[lcore].lcore_ll->dev_removal_flag
987 /* Add the entries back to the lcore and main free ll.*/
988 put_data_ll_free_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_free,
990 put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
992 /* Decrement number of device on the lcore. */
993 lcore_info[vdev->coreid].lcore_ll->device_num--;
995 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been removed "
996 "from data core\n", dev->device_fh);
1003 * A new device is added to a data core. First the device is added
1004 * to the main linked list and the allocated to a specific data core.
1007 new_device(struct virtio_net *dev)
1009 struct virtio_net_data_ll *ll_dev;
1010 int lcore, core_add = 0;
1011 uint32_t device_num_min = nb_devices;
1012 struct vhost_dev *vdev;
1014 vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
1016 RTE_LOG(INFO, VHOST_DATA,
1017 "(%"PRIu64") Couldn't allocate memory for vhost dev\n",
1023 /* Add device to main ll */
1024 ll_dev = get_data_ll_free_entry(&ll_root_free);
1025 if (ll_dev == NULL) {
1026 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in"
1027 " linked list Device limit of %d devices per core"
1028 " has been reached\n", dev->device_fh, nb_devices);
1029 if (vdev->regions_hpa)
1030 rte_free(vdev->regions_hpa);
1034 ll_dev->vdev = vdev;
1035 add_data_ll_entry(&ll_root_used, ll_dev);
1036 vdev->rx_q = dev->device_fh;
1038 /* reset ready flag */
1039 vdev->ready = DEVICE_MAC_LEARNING;
1042 /* Find a suitable lcore to add the device. */
1043 RTE_LCORE_FOREACH_SLAVE(lcore) {
1044 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
1045 device_num_min = lcore_info[lcore].lcore_ll->device_num;
1049 /* Add device to lcore ll */
1050 ll_dev = get_data_ll_free_entry(&lcore_info[core_add].lcore_ll->ll_root_free);
1051 if (ll_dev == NULL) {
1052 RTE_LOG(INFO, VHOST_DATA,
1053 "(%"PRIu64") Failed to add device to data core\n",
1055 vdev->ready = DEVICE_SAFE_REMOVE;
1056 destroy_device(dev);
1057 rte_free(vdev->regions_hpa);
1061 ll_dev->vdev = vdev;
1062 vdev->coreid = core_add;
1064 add_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
1067 /* Initialize device stats */
1068 memset(&dev_statistics[dev->device_fh], 0,
1069 sizeof(struct device_statistics));
1071 /* Disable notifications. */
1072 rte_vhost_enable_guest_notification(dev, VIRTIO_RXQ, 0);
1073 rte_vhost_enable_guest_notification(dev, VIRTIO_TXQ, 0);
1074 lcore_info[vdev->coreid].lcore_ll->device_num++;
1075 dev->flags |= VIRTIO_DEV_RUNNING;
1077 RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n",
1078 dev->device_fh, vdev->coreid);
1084 * These callback allow devices to be added to the data core when configuration
1085 * has been fully complete.
1087 static const struct virtio_net_device_ops virtio_net_device_ops = {
1088 .new_device = new_device,
1089 .destroy_device = destroy_device,
1093 * This is a thread will wake up after a period to print stats if the user has
1099 struct virtio_net_data_ll *dev_ll;
1100 uint64_t tx_dropped, rx_dropped;
1101 uint64_t tx, tx_total, rx, rx_total, rx_ip_csum, rx_l4_csum;
1103 const char clr[] = { 27, '[', '2', 'J', '\0' };
1104 const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
1107 sleep(enable_stats);
1109 /* Clear screen and move to top left */
1110 printf("%s%s", clr, top_left);
1112 printf("\nDevice statistics ================================");
1114 dev_ll = ll_root_used;
1115 while (dev_ll != NULL) {
1116 device_fh = (uint32_t)dev_ll->vdev->dev->device_fh;
1117 tx_total = dev_statistics[device_fh].tx_total;
1118 tx = dev_statistics[device_fh].tx;
1119 tx_dropped = tx_total - tx;
1121 rx_total = rte_atomic64_read(
1122 &dev_statistics[device_fh].rx_total_atomic);
1123 rx = rte_atomic64_read(
1124 &dev_statistics[device_fh].rx_atomic);
1125 rx_dropped = rx_total - rx;
1126 rx_ip_csum = rte_atomic64_read(
1127 &dev_statistics[device_fh].rx_bad_ip_csum);
1128 rx_l4_csum = rte_atomic64_read(
1129 &dev_statistics[device_fh].rx_bad_l4_csum);
1131 printf("\nStatistics for device %"PRIu32" ----------"
1132 "\nTX total: %"PRIu64""
1133 "\nTX dropped: %"PRIu64""
1134 "\nTX successful: %"PRIu64""
1135 "\nRX total: %"PRIu64""
1136 "\nRX bad IP csum: %"PRIu64""
1137 "\nRX bad L4 csum: %"PRIu64""
1138 "\nRX dropped: %"PRIu64""
1139 "\nRX successful: %"PRIu64"",
1150 dev_ll = dev_ll->next;
1152 printf("\n================================================\n");
1157 * Main function, does initialisation and calls the per-lcore functions. The CUSE
1158 * device is also registered here to handle the IOCTLs.
1161 main(int argc, char *argv[])
1163 struct rte_mempool *mbuf_pool = NULL;
1164 unsigned lcore_id, core_id = 0;
1165 unsigned nb_ports, valid_nb_ports;
1169 static pthread_t tid;
1170 char thread_name[RTE_MAX_THREAD_NAME_LEN];
1173 ret = rte_eal_init(argc, argv);
1175 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1179 /* parse app arguments */
1180 ret = tep_termination_parse_args(argc, argv);
1182 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1184 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
1185 if (rte_lcore_is_enabled(lcore_id))
1186 lcore_ids[core_id++] = lcore_id;
1188 /* set the number of swithcing cores available */
1189 nb_switching_cores = rte_lcore_count()-1;
1191 /* Get the number of physical ports. */
1192 nb_ports = rte_eth_dev_count();
1193 if (nb_ports > RTE_MAX_ETHPORTS)
1194 nb_ports = RTE_MAX_ETHPORTS;
1197 * Update the global var NB_PORTS and global array PORTS
1198 * and get value of var VALID_NB_PORTS according to system ports number
1200 valid_nb_ports = check_ports_num(nb_ports);
1202 if ((valid_nb_ports == 0) || (valid_nb_ports > MAX_SUP_PORTS)) {
1203 rte_exit(EXIT_FAILURE, "Current enabled port number is %u,"
1204 "but only %u port can be enabled\n", nb_ports,
1207 /* Create the mbuf pool. */
1208 mbuf_pool = rte_mempool_create(
1212 MBUF_SIZE, MBUF_CACHE_SIZE,
1213 sizeof(struct rte_pktmbuf_pool_private),
1214 rte_pktmbuf_pool_init, NULL,
1215 rte_pktmbuf_init, NULL,
1216 rte_socket_id(), 0);
1217 if (mbuf_pool == NULL)
1218 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1220 for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
1221 vpool_array[queue_id].pool = mbuf_pool;
1223 /* Set log level. */
1224 rte_set_log_level(LOG_LEVEL);
1226 /* initialize all ports */
1227 for (portid = 0; portid < nb_ports; portid++) {
1228 /* skip ports that are not enabled */
1229 if ((enabled_port_mask & (1 << portid)) == 0) {
1230 RTE_LOG(INFO, VHOST_PORT,
1231 "Skipping disabled port %d\n", portid);
1234 if (overlay_options.port_configure(portid, mbuf_pool) != 0)
1235 rte_exit(EXIT_FAILURE,
1236 "Cannot initialize network ports\n");
1239 /* Initialise all linked lists. */
1240 if (init_data_ll() == -1)
1241 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1243 /* Initialize device stats */
1244 memset(&dev_statistics, 0, sizeof(dev_statistics));
1246 /* Enable stats if the user option is set. */
1248 ret = pthread_create(&tid, NULL, (void *)print_stats, NULL);
1250 rte_exit(EXIT_FAILURE, "Cannot create print-stats thread\n");
1251 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN, "print-stats");
1252 ret = rte_thread_setname(tid, thread_name);
1254 RTE_LOG(ERR, VHOST_CONFIG, "Cannot set print-stats name\n");
1257 /* Launch all data cores. */
1258 RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1259 rte_eal_remote_launch(switch_worker,
1260 mbuf_pool, lcore_id);
1262 rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1264 /* Register CUSE device to handle IOCTLs. */
1265 ret = rte_vhost_driver_register((char *)&dev_basename);
1267 rte_exit(EXIT_FAILURE, "CUSE device setup failure.\n");
1269 rte_vhost_driver_callback_register(&virtio_net_device_ops);
1271 /* Start CUSE session. */
1272 rte_vhost_driver_session_start();