/**
* Main init function for the multi-process server app,
* calls subfunctions to do each stage of the initialisation.
*/
int
init(int argc, char *argv[])
{
int retval;
const struct rte_memzone *mz;
uint8_t i, total_ports;
/* init EAL, parsing EAL args */
retval = rte_eal_init(argc, argv);
if (retval < 0)
return -1;
argc -= retval;
argv += retval;
if (rte_eal_pci_probe())
rte_panic("Cannot probe PCI\n");
/* initialise the nic drivers */
retval = init_drivers();
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot initialise drivers\n");
/* get total number of ports */
total_ports = rte_eth_dev_count();
/* set up array for port data */
mz = rte_memzone_reserve(MZ_PORT_INFO, sizeof(*ports),
rte_socket_id(), NO_FLAGS);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot reserve memory zone for port information\n");
memset(mz->addr, 0, sizeof(*ports));
ports = mz->addr;
RTE_LOG(INFO, APP, "memzone address is %lx\n", mz->phys_addr);
/* set up array for statistics */
mz = rte_memzone_reserve(MZ_STATS_INFO, VPORT_STATS_SIZE, rte_socket_id(), NO_FLAGS);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot reserve memory zone for statistics\n");
memset(mz->addr, 0, VPORT_STATS_SIZE);
vport_stats = mz->addr;
/* set up array for flow table data */
mz = rte_memzone_reserve(MZ_FLOW_TABLE, sizeof(*flow_table),
rte_socket_id(), NO_FLAGS);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot reserve memory zone for port information\n");
memset(mz->addr, 0, sizeof(*flow_table));
flow_table = mz->addr;
/* parse additional, application arguments */
retval = parse_app_args(total_ports, argc, argv);
if (retval != 0)
return -1;
/* initialise mbuf pools */
retval = init_mbuf_pools();
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot create needed mbuf pools\n");
/* now initialise the ports we will use */
for (i = 0; i < ports->num_ports; i++) {
retval = init_port(ports->id[i]);
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot initialise port %u\n",
(unsigned)i);
}
/* initialise the client queues/rings for inter process comms */
init_shm_rings();
/* initalise kni queues */
init_kni();
return 0;
}
int
main(int argc, char **argv)
{
int ret;
unsigned nb_ports;
unsigned int lcore_id, last_lcore_id, master_lcore_id;
uint8_t port_id;
uint8_t nb_ports_available;
struct worker_thread_args worker_args = {NULL, NULL};
struct send_thread_args send_args = {NULL, NULL};
struct rte_ring *rx_to_workers;
struct rte_ring *workers_to_tx;
/* catch ctrl-c so we can print on exit */
signal(SIGINT, int_handler);
/* Initialize EAL */
ret = rte_eal_init(argc, argv);
if (ret < 0)
return -1;
argc -= ret;
argv += ret;
/* Parse the application specific arguments */
ret = parse_args(argc, argv);
if (ret < 0)
return -1;
/* Check if we have enought cores */
if (rte_lcore_count() < 3)
rte_exit(EXIT_FAILURE, "Error, This application needs at "
"least 3 logical cores to run:\n"
"1 lcore for packet RX\n"
"1 lcore for packet TX\n"
"and at least 1 lcore for worker threads\n");
nb_ports = rte_eth_dev_count();
if (nb_ports == 0)
rte_exit(EXIT_FAILURE, "Error: no ethernet ports detected\n");
if (nb_ports != 1 && (nb_ports & 1))
rte_exit(EXIT_FAILURE, "Error: number of ports must be even, except "
"when using a single port\n");
mbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", MBUF_PER_POOL,
MBUF_POOL_CACHE_SIZE, 0, MBUF_DATA_SIZE,
rte_socket_id());
if (mbuf_pool == NULL)
rte_exit(EXIT_FAILURE, "%s\n", rte_strerror(rte_errno));
nb_ports_available = nb_ports;
/* initialize all ports */
for (port_id = 0; port_id < nb_ports; port_id++) {
/* skip ports that are not enabled */
if ((portmask & (1 << port_id)) == 0) {
printf("\nSkipping disabled port %d\n", port_id);
nb_ports_available--;
continue;
}
/* init port */
printf("Initializing port %u... done\n", (unsigned) port_id);
if (configure_eth_port(port_id) != 0)
rte_exit(EXIT_FAILURE, "Cannot initialize port %"PRIu8"\n",
port_id);
}
if (!nb_ports_available) {
rte_exit(EXIT_FAILURE,
"All available ports are disabled. Please set portmask.\n");
}
/* Create rings for inter core communication */
rx_to_workers = rte_ring_create("rx_to_workers", RING_SIZE, rte_socket_id(),
RING_F_SP_ENQ);
if (rx_to_workers == NULL)
rte_exit(EXIT_FAILURE, "%s\n", rte_strerror(rte_errno));
workers_to_tx = rte_ring_create("workers_to_tx", RING_SIZE, rte_socket_id(),
RING_F_SC_DEQ);
if (workers_to_tx == NULL)
rte_exit(EXIT_FAILURE, "%s\n", rte_strerror(rte_errno));
if (!disable_reorder) {
send_args.buffer = rte_reorder_create("PKT_RO", rte_socket_id(),
REORDER_BUFFER_SIZE);
if (send_args.buffer == NULL)
rte_exit(EXIT_FAILURE, "%s\n", rte_strerror(rte_errno));
}
last_lcore_id = get_last_lcore_id();
master_lcore_id = rte_get_master_lcore();
worker_args.ring_in = rx_to_workers;
worker_args.ring_out = workers_to_tx;
/* Start worker_thread() on all the available slave cores but the last 1 */
for (lcore_id = 0; lcore_id <= get_previous_lcore_id(last_lcore_id); lcore_id++)
if (rte_lcore_is_enabled(lcore_id) && lcore_id != master_lcore_id)
rte_eal_remote_launch(worker_thread, (void *)&worker_args,
lcore_id);
if (disable_reorder) {
/* Start tx_thread() on the last slave core */
rte_eal_remote_launch((lcore_function_t *)tx_thread, workers_to_tx,
last_lcore_id);
} else {
send_args.ring_in = workers_to_tx;
/* Start send_thread() on the last slave core */
rte_eal_remote_launch((lcore_function_t *)send_thread,
(void *)&send_args, last_lcore_id);
}
/* Start rx_thread() on the master core */
rx_thread(rx_to_workers);
RTE_LCORE_FOREACH_SLAVE(lcore_id) {
if (rte_eal_wait_lcore(lcore_id) < 0)
return -1;
}
print_stats();
return 0;
}
static int
vmxnet3_dev_configure(struct rte_eth_dev *dev)
{
const struct rte_memzone *mz;
struct vmxnet3_hw *hw = dev->data->dev_private;
size_t size;
PMD_INIT_FUNC_TRACE();
if (dev->data->nb_rx_queues > UINT8_MAX ||
dev->data->nb_tx_queues > UINT8_MAX)
return -EINVAL;
size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
if (size > UINT16_MAX)
return -EINVAL;
hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
/*
* Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
* on current socket
*/
mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
"shared", rte_socket_id(), 8);
if (mz == NULL) {
PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
return -ENOMEM;
}
memset(mz->addr, 0, mz->len);
hw->shared = mz->addr;
hw->sharedPA = mz->phys_addr;
/*
* Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
* on current socket
*/
mz = gpa_zone_reserve(dev, size, "queuedesc",
rte_socket_id(), VMXNET3_QUEUE_DESC_ALIGN);
if (mz == NULL) {
PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
return -ENOMEM;
}
memset(mz->addr, 0, mz->len);
hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
hw->queueDescPA = mz->phys_addr;
hw->queue_desc_len = (uint16_t)size;
if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
/* Allocate memory structure for UPT1_RSSConf and configure */
mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf), "rss_conf",
rte_socket_id(), RTE_CACHE_LINE_SIZE);
if (mz == NULL) {
PMD_INIT_LOG(ERR,
"ERROR: Creating rss_conf structure zone");
return -ENOMEM;
}
memset(mz->addr, 0, mz->len);
hw->rss_conf = mz->addr;
hw->rss_confPA = mz->phys_addr;
}
return 0;
}
/* Create a new flow table made of an rte_hash table and a fixed size
* data array for storing values. Only supports IPv4 5-tuple lookups. */
struct onvm_ft*
onvm_ft_create(int cnt, int entry_size) {
struct rte_hash* hash;
struct onvm_ft* ft;
struct rte_hash_parameters ipv4_hash_params = {
.name = NULL,
.entries = cnt,
.key_len = sizeof(struct onvm_ft_ipv4_5tuple),
.hash_func = NULL,
.hash_func_init_val = 0,
};
char s[64];
/* create ipv4 hash table. use core number and cycle counter to get a unique name. */
ipv4_hash_params.name = s;
ipv4_hash_params.socket_id = rte_socket_id();
snprintf(s, sizeof(s), "onvm_ft_%d-%"PRIu64, rte_lcore_id(), rte_get_tsc_cycles());
hash = rte_hash_create(&ipv4_hash_params);
if (hash == NULL) {
return NULL;
}
ft = (struct onvm_ft*)rte_calloc("table", 1, sizeof(struct onvm_ft), 0);
if (ft == NULL) {
rte_hash_free(hash);
return NULL;
}
ft->hash = hash;
ft->cnt = cnt;
ft->entry_size = entry_size;
/* Create data array for storing values */
ft->data = rte_calloc("entry", cnt, entry_size, 0);
if (ft->data == NULL) {
rte_hash_free(hash);
rte_free(ft);
return NULL;
}
return ft;
}
/* Add an entry in flow table and set data to point to the new value.
Returns:
index in the array on success
-EPROTONOSUPPORT if packet is not ipv4.
-EINVAL if the parameters are invalid.
-ENOSPC if there is no space in the hash for this key.
*/
int
onvm_ft_add_pkt(struct onvm_ft* table, struct rte_mbuf *pkt, char** data) {
int32_t tbl_index;
struct onvm_ft_ipv4_5tuple key;
int err;
err = onvm_ft_fill_key(&key, pkt);
if (err < 0) {
return err;
}
tbl_index = rte_hash_add_key_with_hash(table->hash, (const void *)&key, pkt->hash.rss);
if (tbl_index >= 0) {
*data = &table->data[tbl_index*table->entry_size];
}
return tbl_index;
}
static int
test_distributor_perf(void)
{
static struct rte_distributor *ds;
static struct rte_distributor *db;
static struct rte_mempool *p;
if (rte_lcore_count() < 2) {
printf("ERROR: not enough cores to test distributor\n");
return -1;
}
/* first time how long it takes to round-trip a cache line */
time_cache_line_switch();
if (ds == NULL) {
ds = rte_distributor_create("Test_perf", rte_socket_id(),
rte_lcore_count() - 1,
RTE_DIST_ALG_SINGLE);
if (ds == NULL) {
printf("Error creating distributor\n");
return -1;
}
} else {
rte_distributor_clear_returns(ds);
}
if (db == NULL) {
db = rte_distributor_create("Test_burst", rte_socket_id(),
rte_lcore_count() - 1,
RTE_DIST_ALG_BURST);
if (db == NULL) {
printf("Error creating burst distributor\n");
return -1;
}
} else {
rte_distributor_clear_returns(db);
}
const unsigned nb_bufs = (511 * rte_lcore_count()) < BIG_BATCH ?
(BIG_BATCH * 2) - 1 : (511 * rte_lcore_count());
if (p == NULL) {
p = rte_pktmbuf_pool_create("DPT_MBUF_POOL", nb_bufs, BURST,
0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
if (p == NULL) {
printf("Error creating mempool\n");
return -1;
}
}
printf("=== Performance test of distributor (single mode) ===\n");
rte_eal_mp_remote_launch(handle_work, ds, SKIP_MASTER);
if (perf_test(ds, p) < 0)
return -1;
quit_workers(ds, p);
printf("=== Performance test of distributor (burst mode) ===\n");
rte_eal_mp_remote_launch(handle_work, db, SKIP_MASTER);
if (perf_test(db, p) < 0)
return -1;
quit_workers(db, p);
return 0;
}
int main(int argc, char *argv[])
{
int err, ret;
uint32_t i, pmsk;
struct nmreq req;
struct pollfd pollfd[MAX_PORT_NUM];
struct rte_mempool *pool;
struct netmap_ring *rx_ring, *tx_ring;
ret = rte_eal_init(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Cannot initialize EAL\n");
argc -= ret;
argv += ret;
parse_args(argc, argv);
if (ports.num == 0)
rte_exit(EXIT_FAILURE, "no ports specified\n");
if (rte_eth_dev_count() < 1)
rte_exit(EXIT_FAILURE, "Not enough ethernet ports available\n");
pool = rte_pktmbuf_pool_create("mbuf_pool", MBUF_PER_POOL, 32, 0,
MBUF_DATA_SIZE, rte_socket_id());
if (pool == NULL)
rte_exit(EXIT_FAILURE, "Couldn't create mempool\n");
netmap_conf.socket_id = rte_socket_id();
err = rte_netmap_init(&netmap_conf);
if (err < 0)
rte_exit(EXIT_FAILURE,
"Couldn't initialize librte_compat_netmap\n");
else
printf("librte_compat_netmap initialized\n");
port_conf.pool = pool;
port_conf.socket_id = rte_socket_id();
for (i = 0; i != ports.num; i++) {
err = rte_netmap_init_port(ports.p[i].id, &port_conf);
if (err < 0)
rte_exit(EXIT_FAILURE, "Couldn't setup port %hhu\n",
ports.p[i].id);
rte_eth_promiscuous_enable(ports.p[i].id);
}
for (i = 0; i != ports.num; i++) {
err = netmap_port_open(i);
if (err) {
rte_exit(EXIT_FAILURE, "Couldn't set port %hhu "
"under NETMAP control\n",
ports.p[i].id);
}
else
printf("Port %hhu now in Netmap mode\n", ports.p[i].id);
}
memset(pollfd, 0, sizeof(pollfd));
for (i = 0; i != ports.num; i++) {
pollfd[i].fd = ports.p[i].fd;
pollfd[i].events = POLLIN | POLLOUT;
}
signal(SIGINT, sigint_handler);
pmsk = ports.num - 1;
printf("Bridge up and running!\n");
while (!stop) {
uint32_t n_pkts;
pollfd[0].revents = 0;
pollfd[1].revents = 0;
ret = rte_netmap_poll(pollfd, ports.num, 0);
if (ret < 0) {
stop = 1;
printf("[E] poll returned with error %d\n", ret);
}
if (((pollfd[0].revents | pollfd[1].revents) & POLLERR) != 0) {
printf("POLLERR!\n");
}
if ((pollfd[0].revents & POLLIN) != 0 &&
(pollfd[pmsk].revents & POLLOUT) != 0) {
rx_ring = ports.p[0].rx_ring;
tx_ring = ports.p[pmsk].tx_ring;
n_pkts = RTE_MIN(rx_ring->avail, tx_ring->avail);
move(n_pkts, rx_ring, tx_ring);
}
if (pmsk != 0 && (pollfd[pmsk].revents & POLLIN) != 0 &&
(pollfd[0].revents & POLLOUT) != 0) {
rx_ring = ports.p[pmsk].rx_ring;
tx_ring = ports.p[0].tx_ring;
n_pkts = RTE_MIN(rx_ring->avail, tx_ring->avail);
move(n_pkts, rx_ring, tx_ring);
}
}
printf("Bridge stopped!\n");
for (i = 0; i != ports.num; i++) {
err = rte_netmap_ioctl(ports.p[i].fd, NIOCUNREGIF, &req);
if (err) {
printf("[E] NIOCUNREGIF ioctl failed (error %d)\n",
err);
}
else
printf("Port %hhu unregistered from Netmap mode\n", ports.p[i].id);
rte_netmap_close(ports.p[i].fd);
}
return 0;
}
void
app_main_loop_pipeline_passthrough(void) {
struct rte_pipeline_params pipeline_params = {
.name = "pipeline",
.socket_id = rte_socket_id(),
};
struct rte_pipeline *p;
uint32_t port_in_id[APP_MAX_PORTS];
uint32_t port_out_id[APP_MAX_PORTS];
uint32_t table_id[APP_MAX_PORTS];
uint32_t i;
uint32_t core_id = rte_lcore_id();
struct app_core_params *core_params = app_get_core_params(core_id);
if ((core_params == NULL) || (core_params->core_type != APP_CORE_PT))
rte_panic("Core %u misconfiguration\n", core_id);
RTE_LOG(INFO, USER1, "Core %u is doing pass-through\n", core_id);
/* Pipeline configuration */
p = rte_pipeline_create(&pipeline_params);
if (p == NULL)
rte_panic("%s: Unable to configure the pipeline\n", __func__);
/* Input port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_reader_params port_ring_params = {
.ring = app.rings[core_params->swq_in[i]],
};
struct rte_pipeline_port_in_params port_params = {
.ops = &rte_port_ring_reader_ops,
.arg_create = (void *) &port_ring_params,
.f_action = NULL,
.arg_ah = NULL,
.burst_size = app.bsz_swq_rd,
};
if (rte_pipeline_port_in_create(p, &port_params,
&port_in_id[i])) {
rte_panic("%s: Unable to configure input port for "
"ring %d\n", __func__, i);
}
}
/* Output port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_writer_params port_ring_params = {
.ring = app.rings[core_params->swq_out[i]],
.tx_burst_sz = app.bsz_swq_wr,
};
struct rte_pipeline_port_out_params port_params = {
.ops = &rte_port_ring_writer_ops,
.arg_create = (void *) &port_ring_params,
.f_action = NULL,
.f_action_bulk = NULL,
.arg_ah = NULL,
};
if (rte_pipeline_port_out_create(p, &port_params,
&port_out_id[i])) {
rte_panic("%s: Unable to configure output port for "
"ring %d\n", __func__, i);
}
}
/* Table configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_pipeline_table_params table_params = {
.ops = &rte_table_stub_ops,
.arg_create = NULL,
.f_action_hit = NULL,
.f_action_miss = NULL,
.arg_ah = NULL,
.action_data_size = 0,
};
if (rte_pipeline_table_create(p, &table_params, &table_id[i]))
rte_panic("%s: Unable to configure table %u\n",
__func__, i);
}
/* Interconnecting ports and tables */
for (i = 0; i < app.n_ports; i++) {
if (rte_pipeline_port_in_connect_to_table(p, port_in_id[i],
table_id[i])) {
rte_panic("%s: Unable to connect input port %u to "
"table %u\n", __func__, port_in_id[i],
table_id[i]);
}
}
/* Add entries to tables */
for (i = 0; i < app.n_ports; i++) {
struct rte_pipeline_table_entry default_entry = {
.action = RTE_PIPELINE_ACTION_PORT,
{.port_id = port_out_id[i]},
};
struct rte_pipeline_table_entry *default_entry_ptr;
if (rte_pipeline_table_default_entry_add(p, table_id[i],
&default_entry, &default_entry_ptr))
rte_panic("%s: Unable to add default entry to "
"table %u\n", __func__, table_id[i]);
}
/* Enable input ports */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_enable(p, port_in_id[i]))
rte_panic("Unable to enable input port %u\n",
port_in_id[i]);
/* Check pipeline consistency */
if (rte_pipeline_check(p) < 0)
rte_panic("%s: Pipeline consistency check failed\n", __func__);
/* Run-time */
for (i = 0; ; i++) {
rte_pipeline_run(p);
if ((i & APP_FLUSH) == 0)
rte_pipeline_flush(p);
}
}
void
app_main_loop_passthrough(void) {
struct app_mbuf_array *m;
uint32_t i;
uint32_t core_id = rte_lcore_id();
struct app_core_params *core_params = app_get_core_params(core_id);
if ((core_params == NULL) || (core_params->core_type != APP_CORE_PT))
rte_panic("Core %u misconfiguration\n", core_id);
RTE_LOG(INFO, USER1, "Core %u is doing pass-through (no pipeline)\n",
core_id);
m = rte_malloc_socket(NULL, sizeof(struct app_mbuf_array),
RTE_CACHE_LINE_SIZE, rte_socket_id());
if (m == NULL)
rte_panic("%s: cannot allocate buffer space\n", __func__);
for (i = 0; ; i = ((i + 1) & (app.n_ports - 1))) {
int ret;
ret = rte_ring_sc_dequeue_bulk(
app.rings[core_params->swq_in[i]],
(void **) m->array,
app.bsz_swq_rd);
if (ret == -ENOENT)
continue;
do {
ret = rte_ring_sp_enqueue_bulk(
app.rings[core_params->swq_out[i]],
(void **) m->array,
app.bsz_swq_wr);
} while (ret < 0);
}
}
static int
fs_eth_dev_create(struct rte_vdev_device *vdev)
{
struct rte_eth_dev *dev;
struct ether_addr *mac;
struct fs_priv *priv;
struct sub_device *sdev;
const char *params;
unsigned int socket_id;
uint8_t i;
int ret;
dev = NULL;
priv = NULL;
socket_id = rte_socket_id();
INFO("Creating fail-safe device on NUMA socket %u", socket_id);
params = rte_vdev_device_args(vdev);
if (params == NULL) {
ERROR("This PMD requires sub-devices, none provided");
return -1;
}
dev = rte_eth_vdev_allocate(vdev, sizeof(*priv));
if (dev == NULL) {
ERROR("Unable to allocate rte_eth_dev");
return -1;
}
priv = PRIV(dev);
priv->dev = dev;
dev->dev_ops = &failsafe_ops;
dev->data->mac_addrs = &PRIV(dev)->mac_addrs[0];
dev->data->dev_link = eth_link;
PRIV(dev)->nb_mac_addr = 1;
TAILQ_INIT(&PRIV(dev)->flow_list);
dev->rx_pkt_burst = (eth_rx_burst_t)&failsafe_rx_burst;
dev->tx_pkt_burst = (eth_tx_burst_t)&failsafe_tx_burst;
ret = fs_sub_device_alloc(dev, params);
if (ret) {
ERROR("Could not allocate sub_devices");
goto free_dev;
}
ret = failsafe_args_parse(dev, params);
if (ret)
goto free_subs;
ret = rte_eth_dev_owner_new(&priv->my_owner.id);
if (ret) {
ERROR("Failed to get unique owner identifier");
goto free_args;
}
snprintf(priv->my_owner.name, sizeof(priv->my_owner.name),
FAILSAFE_OWNER_NAME);
DEBUG("Failsafe port %u owner info: %s_%016"PRIX64, dev->data->port_id,
priv->my_owner.name, priv->my_owner.id);
ret = rte_eth_dev_callback_register(RTE_ETH_ALL, RTE_ETH_EVENT_NEW,
failsafe_eth_new_event_callback,
dev);
if (ret) {
ERROR("Failed to register NEW callback");
goto free_args;
}
ret = failsafe_eal_init(dev);
if (ret)
goto unregister_new_callback;
ret = fs_mutex_init(priv);
if (ret)
goto unregister_new_callback;
ret = failsafe_hotplug_alarm_install(dev);
if (ret) {
ERROR("Could not set up plug-in event detection");
goto unregister_new_callback;
}
mac = &dev->data->mac_addrs[0];
if (failsafe_mac_from_arg) {
/*
* If MAC address was provided as a parameter,
* apply to all probed slaves.
*/
FOREACH_SUBDEV_STATE(sdev, i, dev, DEV_PROBED) {
ret = rte_eth_dev_default_mac_addr_set(PORT_ID(sdev),
mac);
if (ret) {
ERROR("Failed to set default MAC address");
goto cancel_alarm;
}
}
} else {
/* Main function */
int main(int argc, char **argv)
{
int ret;
int i;
/* Create handler for SIGINT for CTRL + C closing and SIGALRM to print stats*/
signal(SIGINT, sig_handler);
signal(SIGALRM, alarm_routine);
/* Initialize DPDK enviroment with args, then shift argc and argv to get application parameters */
ret = rte_eal_init(argc, argv);
if (ret < 0) FATAL_ERROR("Cannot init EAL\n");
argc -= ret;
argv += ret;
/* Check if this application can use two cores*/
ret = rte_lcore_count ();
if (ret != 2) PRINT_INFO("This application needs exactly two (2) cores.");
/* Parse arguments */
parse_args(argc, argv);
if (ret < 0) FATAL_ERROR("Wrong arguments\n");
/* Probe PCI bus for ethernet devices, mandatory only in DPDK < 1.8.0 */
#if RTE_VER_MAJOR == 1 && RTE_VER_MINOR < 8
ret = rte_eal_pci_probe();
if (ret < 0) FATAL_ERROR("Cannot probe PCI\n");
#endif
/* Get number of ethernet devices */
nb_sys_ports = rte_eth_dev_count();
//if (nb_sys_ports <= 0) FATAL_ERROR("Cannot find ETH devices\n");
/* Create a mempool with per-core cache, initializing every element for be used as mbuf, and allocating on the current NUMA node */
// pktmbuf_pool = rte_mempool_create(MEMPOOL_NAME, buffer_size-1, MEMPOOL_ELEM_SZ, MEMPOOL_CACHE_SZ, sizeof(struct rte_pktmbuf_pool_private), rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,rte_socket_id(), 0);
pktmbuf_pool = rte_pktmbuf_pool_create(MEMPOOL_NAME,buffer_size-1, MEMPOOL_CACHE_SZ, 0, snaplen + RTE_PKTMBUF_HEADROOM, rte_socket_id());
if (pktmbuf_pool == NULL) FATAL_ERROR("Cannot create cluster_mem_pool. Errno: %d [ENOMEM: %d, ENOSPC: %d, E_RTE_NO_TAILQ: %d, E_RTE_NO_CONFIG: %d, E_RTE_SECONDARY: %d, EINVAL: %d, EEXIST: %d]\n", rte_errno, ENOMEM, ENOSPC, RTE_MAX_TAILQ/*E_RTE_NO_TAILQ*/, E_RTE_NO_CONFIG, E_RTE_SECONDARY, EINVAL, EEXIST );
/* Init intermediate queue data structures: the ring. */
intermediate_ring = rte_ring_create (INTERMEDIATERING_NAME, buffer_size, rte_socket_id(), RING_F_SP_ENQ | RING_F_SC_DEQ );
if (intermediate_ring == NULL ) FATAL_ERROR("Cannot create ring");
/* Operations needed for each ethernet device */
for(i=0; i < nb_sys_ports; i++)
init_port(i);
/* Start Pcap */
//PcapStartUp();
/* Start consumer and producer routine on 2 different cores: consumer launched first... */
ret = rte_eal_mp_remote_launch (main_loop_producer, NULL, SKIP_MASTER);
if (ret != 0) FATAL_ERROR("Cannot start consumer thread\n");
/*RTE_LCORE_FOREACH_SLAVE(i) {
if (rte_eal_wait_lcore(i) < 0)
return -1;
}*/
/* ... and then loop in consumer */
//main_loop_producer ( NULL );
main_loop_consumer(NULL);
return 0;
}
int
test_rw_piot(int argc, char **argv)
{
int ret;
rw_piot_open_request_info_t req_param;
rw_piot_open_response_info_t rsp_param;
char devname[256];
/* Init PIOT */
ret = rw_piot_init(argc, argv, (void *)&my_instance, NULL);
ASSERT(ret>=0);
printf("PIOT Init Successful\n");
idx = rw_piot_dev_count();
/* No devices opened yet */
ASSERT(idx == 0);
/* Open first Ring device */
bzero(&req_param, sizeof(req_param));
req_param.num_rx_queues = 1;
req_param.num_tx_queues = 1;
strcpy(devname, "eth_ring:name=ring0");
piot_handle[idx] = rw_piot_open(devname, &req_param, &rsp_param);
ASSERT(piot_handle[idx] != 0);
ASSERT(rw_piot_dev_count() == 1);
printf("PIOT device open %s - Successful\n", devname);
/* Open second Ring device */
idx ++;
bzero(&req_param, sizeof(req_param));
req_param.num_rx_queues = 4;
req_param.num_tx_queues = 4;
strcpy(devname, "eth_ring:name=ring1");
piot_handle[idx] = rw_piot_open(devname, &req_param, &rsp_param);
ASSERT(piot_handle[idx] != 0);
ASSERT(rw_piot_dev_count() == 2);
printf("PIOT device open %s - Successful\n", devname);
/* Open third Ring device */
idx ++;
bzero(&req_param, sizeof(req_param));
req_param.num_rx_queues = 16;
req_param.num_tx_queues = 16;
strcpy(devname, "eth_ring:name=ring2");
piot_handle[idx] = rw_piot_open(devname, &req_param, &rsp_param);
ASSERT(piot_handle[idx] != 0);
ASSERT(rw_piot_dev_count() == 3);
printf("PIOT device open %s - Successful\n", devname);
mp = rte_mempool_create("mbuf_pool", NB_MBUF,
MBUF_SIZE, 32,
sizeof(struct rte_pktmbuf_pool_private),
rte_pktmbuf_pool_init, NULL,
rte_pktmbuf_init, NULL,
rte_socket_id(), 0);
ASSERT(mp);
ASSERT(test_rw_piot_setup_queues() >=0);
printf("PIOT Queue setup Successful\n");
ret = rw_piot_dev_start(piot_handle[0]);
ASSERT(ret == 0);
ret = rw_piot_dev_start(piot_handle[1]);
ASSERT(ret == 0);
ret = rw_piot_dev_start(piot_handle[2]);
ASSERT(ret == 0);
printf("PIOT device start Successful\n");
ASSERT(test_rw_piot_send_packets() >= 0);
printf("PIOT RX/TX - Successful\n");
ASSERT(test_get_stats() >= 0);
printf("PIOT Get Stats - Successful\n");
ASSERT(test_stats_reset() >= 0);
printf("PIOT Reset Stats - Successful\n");
rw_piot_dev_stop(piot_handle[0]);
rw_piot_dev_stop(piot_handle[1]);
rw_piot_dev_stop(piot_handle[2]);
printf("PIOT Stopping Devices - Successful\n");
return 0;
}
static int
fbk_hash_perf_test(void)
{
struct rte_fbk_hash_params params = {
.name = "fbk_hash_test",
.entries = ENTRIES,
.entries_per_bucket = 4,
.socket_id = rte_socket_id(),
};
struct rte_fbk_hash_table *handle = NULL;
uint32_t *keys = NULL;
unsigned indexes[TEST_SIZE];
uint64_t lookup_time = 0;
unsigned added = 0;
unsigned value = 0;
uint32_t key;
uint16_t val;
unsigned i, j;
handle = rte_fbk_hash_create(¶ms);
if (handle == NULL) {
printf("Error creating table\n");
return -1;
}
keys = rte_zmalloc(NULL, ENTRIES * sizeof(*keys), 0);
if (keys == NULL) {
printf("fbk hash: memory allocation for key store failed\n");
return -1;
}
/* Generate random keys and values. */
for (i = 0; i < ENTRIES; i++) {
key = (uint32_t)rte_rand();
key = ((uint64_t)key << 32) | (uint64_t)rte_rand();
val = (uint16_t)rte_rand();
if (rte_fbk_hash_add_key(handle, key, val) == 0) {
keys[added] = key;
added++;
}
if (added > (LOAD_FACTOR * ENTRIES))
break;
}
for (i = 0; i < TEST_ITERATIONS; i++) {
uint64_t begin;
uint64_t end;
/* Generate random indexes into keys[] array. */
for (j = 0; j < TEST_SIZE; j++)
indexes[j] = rte_rand() % added;
begin = rte_rdtsc();
/* Do lookups */
for (j = 0; j < TEST_SIZE; j++)
value += rte_fbk_hash_lookup(handle, keys[indexes[j]]);
end = rte_rdtsc();
lookup_time += (double)(end - begin);
}
printf("\n\n *** FBK Hash function performance test results ***\n");
/*
* The use of the 'value' variable ensures that the hash lookup is not
* being optimised out by the compiler.
*/
if (value != 0)
printf("Number of ticks per lookup = %g\n",
(double)lookup_time /
((double)TEST_ITERATIONS * (double)TEST_SIZE));
rte_fbk_hash_free(handle);
return 0;
}
static int
test_hash_perf(void)
{
unsigned with_pushes;
for (with_pushes = 0; with_pushes <= 1; with_pushes++) {
if (with_pushes == 0)
printf("\nALL ELEMENTS IN PRIMARY LOCATION\n");
else
printf("\nELEMENTS IN PRIMARY OR SECONDARY LOCATION\n");
if (run_all_tbl_perf_tests(with_pushes) < 0)
return -1;
}
if (fbk_hash_perf_test() < 0)
return -1;
return 0;
}
/*
* Initialize a given port using default settings and with the RX buffers
* coming from the mbuf_pool passed as a parameter.
* FIXME: Starting with assumption of one thread/core per port
*/
static inline int uhd_dpdk_port_init(struct uhd_dpdk_port *port,
struct rte_mempool *rx_mbuf_pool,
unsigned int mtu)
{
int retval;
/* Check for a valid port */
if (port->id >= rte_eth_dev_count())
return -ENODEV;
/* Set up Ethernet device with defaults (1 RX ring, 1 TX ring) */
/* FIXME: Check if hw_ip_checksum is possible */
struct rte_eth_conf port_conf = {
.rxmode = {
.max_rx_pkt_len = mtu,
.jumbo_frame = 1,
.hw_ip_checksum = 1,
}
};
retval = rte_eth_dev_configure(port->id, 1, 1, &port_conf);
if (retval != 0)
return retval;
retval = rte_eth_rx_queue_setup(port->id, 0, DEFAULT_RING_SIZE,
rte_eth_dev_socket_id(port->id), NULL, rx_mbuf_pool);
if (retval < 0)
return retval;
retval = rte_eth_tx_queue_setup(port->id, 0, DEFAULT_RING_SIZE,
rte_eth_dev_socket_id(port->id), NULL);
if (retval < 0)
goto port_init_fail;
/* Create the hash table for the RX sockets */
char name[32];
snprintf(name, sizeof(name), "rx_table_%u", port->id);
struct rte_hash_parameters hash_params = {
.name = name,
.entries = UHD_DPDK_MAX_SOCKET_CNT,
.key_len = sizeof(struct uhd_dpdk_ipv4_5tuple),
.hash_func = NULL,
.hash_func_init_val = 0,
};
port->rx_table = rte_hash_create(&hash_params);
if (port->rx_table == NULL) {
retval = rte_errno;
goto port_init_fail;
}
/* Create ARP table */
snprintf(name, sizeof(name), "arp_table_%u", port->id);
hash_params.name = name;
hash_params.entries = UHD_DPDK_MAX_SOCKET_CNT;
hash_params.key_len = sizeof(uint32_t);
hash_params.hash_func = NULL;
hash_params.hash_func_init_val = 0;
port->arp_table = rte_hash_create(&hash_params);
if (port->arp_table == NULL) {
retval = rte_errno;
goto free_rx_table;
}
/* Set up list for TX queues */
LIST_INIT(&port->txq_list);
/* Start the Ethernet port. */
retval = rte_eth_dev_start(port->id);
if (retval < 0) {
goto free_arp_table;
}
/* Display the port MAC address. */
rte_eth_macaddr_get(port->id, &port->mac_addr);
RTE_LOG(INFO, EAL, "Port %u MAC: %02x %02x %02x %02x %02x %02x\n",
(unsigned)port->id,
port->mac_addr.addr_bytes[0], port->mac_addr.addr_bytes[1],
port->mac_addr.addr_bytes[2], port->mac_addr.addr_bytes[3],
port->mac_addr.addr_bytes[4], port->mac_addr.addr_bytes[5]);
struct rte_eth_link link;
rte_eth_link_get(port->id, &link);
RTE_LOG(INFO, EAL, "Port %u UP: %d\n", port->id, link.link_status);
return 0;
free_arp_table:
rte_hash_free(port->arp_table);
free_rx_table:
rte_hash_free(port->rx_table);
port_init_fail:
return rte_errno;
}
static int uhd_dpdk_thread_init(struct uhd_dpdk_thread *thread, unsigned int id)
{
if (!ctx || !thread)
return -EINVAL;
unsigned int socket_id = rte_lcore_to_socket_id(id);
thread->id = id;
thread->rx_pktbuf_pool = ctx->rx_pktbuf_pools[socket_id];
thread->tx_pktbuf_pool = ctx->tx_pktbuf_pools[socket_id];
LIST_INIT(&thread->port_list);
char name[32];
snprintf(name, sizeof(name), "sockreq_ring_%u", id);
thread->sock_req_ring = rte_ring_create(
name,
UHD_DPDK_MAX_PENDING_SOCK_REQS,
socket_id,
RING_F_SC_DEQ
);
if (!thread->sock_req_ring)
return -ENOMEM;
return 0;
}
int uhd_dpdk_init(int argc, char **argv, unsigned int num_ports,
int *port_thread_mapping, int num_mbufs, int mbuf_cache_size,
int mtu)
{
/* Init context only once */
if (ctx)
return 1;
if ((num_ports == 0) || (port_thread_mapping == NULL)) {
return -EINVAL;
}
/* Grabs arguments intended for DPDK's EAL */
int ret = rte_eal_init(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
ctx = (struct uhd_dpdk_ctx *) rte_zmalloc("uhd_dpdk_ctx", sizeof(*ctx), rte_socket_id());
if (!ctx)
return -ENOMEM;
ctx->num_threads = rte_lcore_count();
if (ctx->num_threads <= 1)
rte_exit(EXIT_FAILURE, "Error: No worker threads enabled\n");
/* Check that we have ports to send/receive on */
ctx->num_ports = rte_eth_dev_count();
if (ctx->num_ports < 1)
rte_exit(EXIT_FAILURE, "Error: Found no ports\n");
if (ctx->num_ports < num_ports)
rte_exit(EXIT_FAILURE, "Error: User requested more ports than available\n");
/* Get memory for thread and port data structures */
ctx->threads = rte_zmalloc("uhd_dpdk_thread", RTE_MAX_LCORE*sizeof(struct uhd_dpdk_thread), 0);
if (!ctx->threads)
rte_exit(EXIT_FAILURE, "Error: Could not allocate memory for thread data\n");
ctx->ports = rte_zmalloc("uhd_dpdk_port", ctx->num_ports*sizeof(struct uhd_dpdk_port), 0);
if (!ctx->ports)
rte_exit(EXIT_FAILURE, "Error: Could not allocate memory for port data\n");
/* Initialize the thread data structures */
for (int i = rte_get_next_lcore(-1, 1, 0);
(i < RTE_MAX_LCORE);
i = rte_get_next_lcore(i, 1, 0))
{
/* Do one mempool of RX/TX per socket */
unsigned int socket_id = rte_lcore_to_socket_id(i);
/* FIXME Probably want to take into account actual number of ports per socket */
if (ctx->tx_pktbuf_pools[socket_id] == NULL) {
/* Creates a new mempool in memory to hold the mbufs.
* This is done for each CPU socket
*/
const int mbuf_size = mtu + 2048 + RTE_PKTMBUF_HEADROOM;
char name[32];
snprintf(name, sizeof(name), "rx_mbuf_pool_%u", socket_id);
ctx->rx_pktbuf_pools[socket_id] = rte_pktmbuf_pool_create(
name,
ctx->num_ports*num_mbufs,
mbuf_cache_size,
0,
mbuf_size,
socket_id
);
snprintf(name, sizeof(name), "tx_mbuf_pool_%u", socket_id);
ctx->tx_pktbuf_pools[socket_id] = rte_pktmbuf_pool_create(
name,
ctx->num_ports*num_mbufs,
mbuf_cache_size,
0,
mbuf_size,
socket_id
);
if ((ctx->rx_pktbuf_pools[socket_id]== NULL) ||
(ctx->tx_pktbuf_pools[socket_id]== NULL))
rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
}
if (uhd_dpdk_thread_init(&ctx->threads[i], i) < 0)
rte_exit(EXIT_FAILURE, "Error initializing thread %i\n", i);
}
unsigned master_lcore = rte_get_master_lcore();
/* Assign ports to threads and initialize the port data structures */
for (unsigned int i = 0; i < num_ports; i++) {
int thread_id = port_thread_mapping[i];
if (thread_id < 0)
continue;
if (((unsigned int) thread_id) == master_lcore)
RTE_LOG(WARNING, EAL, "User requested master lcore for port %u\n", i);
if (ctx->threads[thread_id].id != (unsigned int) thread_id)
rte_exit(EXIT_FAILURE, "Requested inactive lcore %u for port %u\n", (unsigned int) thread_id, i);
struct uhd_dpdk_port *port = &ctx->ports[i];
port->id = i;
port->parent = &ctx->threads[thread_id];
ctx->threads[thread_id].num_ports++;
LIST_INSERT_HEAD(&ctx->threads[thread_id].port_list, port, port_entry);
/* Initialize port. */
if (uhd_dpdk_port_init(port, port->parent->rx_pktbuf_pool, mtu) != 0)
rte_exit(EXIT_FAILURE, "Cannot init port %"PRIu8 "\n",
i);
}
RTE_LOG(INFO, EAL, "Init DONE!\n");
/* FIXME: Create functions to do this */
RTE_LOG(INFO, EAL, "Starting I/O threads!\n");
for (int i = rte_get_next_lcore(-1, 1, 0);
(i < RTE_MAX_LCORE);
i = rte_get_next_lcore(i, 1, 0))
{
struct uhd_dpdk_thread *t = &ctx->threads[i];
if (!LIST_EMPTY(&t->port_list)) {
rte_eal_remote_launch(_uhd_dpdk_driver_main, NULL, ctx->threads[i].id);
}
}
return 0;
}
/* FIXME: This will be changed once we have functions to handle the threads */
int uhd_dpdk_destroy(void)
{
if (!ctx)
return -ENODEV;
struct uhd_dpdk_config_req *req = (struct uhd_dpdk_config_req *) rte_zmalloc(NULL, sizeof(*req), 0);
if (!req)
return -ENOMEM;
req->req_type = UHD_DPDK_LCORE_TERM;
for (int i = rte_get_next_lcore(-1, 1, 0);
(i < RTE_MAX_LCORE);
i = rte_get_next_lcore(i, 1, 0))
{
struct uhd_dpdk_thread *t = &ctx->threads[i];
if (LIST_EMPTY(&t->port_list))
continue;
if (rte_eal_get_lcore_state(t->id) == FINISHED)
continue;
pthread_mutex_init(&req->mutex, NULL);
pthread_cond_init(&req->cond, NULL);
pthread_mutex_lock(&req->mutex);
if (rte_ring_enqueue(t->sock_req_ring, req)) {
pthread_mutex_unlock(&req->mutex);
RTE_LOG(ERR, USER2, "Failed to terminate thread %d\n", i);
rte_free(req);
return -ENOSPC;
}
struct timespec timeout = {
.tv_sec = 1,
.tv_nsec = 0
};
pthread_cond_timedwait(&req->cond, &req->mutex, &timeout);
pthread_mutex_unlock(&req->mutex);
}
rte_free(req);
return 0;
}
void
app_main_loop_worker_pipeline_lpm_ipv6(void) {
struct rte_pipeline_params pipeline_params = {
.name = "pipeline",
.socket_id = rte_socket_id(),
};
struct rte_pipeline *p;
uint32_t port_in_id[APP_MAX_PORTS];
uint32_t port_out_id[APP_MAX_PORTS];
uint32_t table_id;
uint32_t i;
RTE_LOG(INFO, USER1,
"Core %u is doing work (pipeline with IPv6 LPM table)\n",
rte_lcore_id());
/* Pipeline configuration */
p = rte_pipeline_create(&pipeline_params);
if (p == NULL)
rte_panic("Unable to configure the pipeline\n");
/* Input port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_reader_params port_ring_params = {
.ring = app.rings_rx[i],
};
struct rte_pipeline_port_in_params port_params = {
.ops = &rte_port_ring_reader_ops,
.arg_create = (void *) &port_ring_params,
.f_action = NULL,
.arg_ah = NULL,
.burst_size = app.burst_size_worker_read,
};
if (rte_pipeline_port_in_create(p, &port_params,
&port_in_id[i]))
rte_panic("Unable to configure input port for "
"ring %d\n", i);
}
/* Output port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_writer_params port_ring_params = {
.ring = app.rings_tx[i],
.tx_burst_sz = app.burst_size_worker_write,
};
struct rte_pipeline_port_out_params port_params = {
.ops = &rte_port_ring_writer_ops,
.arg_create = (void *) &port_ring_params,
.f_action = NULL,
.f_action_bulk = NULL,
.arg_ah = NULL,
};
if (rte_pipeline_port_out_create(p, &port_params,
&port_out_id[i]))
rte_panic("Unable to configure output port for "
"ring %d\n", i);
}
/* Table configuration */
{
struct rte_table_lpm_ipv6_params table_lpm_ipv6_params = {
.name = "LPM",
.n_rules = 1 << 24,
.number_tbl8s = 1 << 21,
.entry_unique_size =
sizeof(struct rte_pipeline_table_entry),
.offset = APP_METADATA_OFFSET(32),
};
struct rte_pipeline_table_params table_params = {
.ops = &rte_table_lpm_ipv6_ops,
.arg_create = &table_lpm_ipv6_params,
.f_action_hit = NULL,
.f_action_miss = NULL,
.arg_ah = NULL,
.action_data_size = 0,
};
if (rte_pipeline_table_create(p, &table_params, &table_id))
rte_panic("Unable to configure the IPv6 LPM table\n");
}
/* Interconnecting ports and tables */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_connect_to_table(p, port_in_id[i],
table_id))
rte_panic("Unable to connect input port %u to "
"table %u\n", port_in_id[i], table_id);
/* Add entries to tables */
for (i = 0; i < app.n_ports; i++) {
struct rte_pipeline_table_entry entry = {
.action = RTE_PIPELINE_ACTION_PORT,
{.port_id = port_out_id[i & (app.n_ports - 1)]},
};
struct rte_table_lpm_ipv6_key key;
struct rte_pipeline_table_entry *entry_ptr;
uint32_t ip;
int key_found, status;
key.depth = 8 + __builtin_popcount(app.n_ports - 1);
ip = rte_bswap32(i << (24 -
__builtin_popcount(app.n_ports - 1)));
memcpy(key.ip, &ip, sizeof(uint32_t));
printf("Adding rule to IPv6 LPM table (IPv6 destination = "
"%.2x%.2x:%.2x%.2x:%.2x%.2x:%.2x%.2x:"
"%.2x%.2x:%.2x%.2x:%.2x%.2x:%.2x%.2x/%u => "
"port out = %u)\n",
key.ip[0], key.ip[1], key.ip[2], key.ip[3],
key.ip[4], key.ip[5], key.ip[6], key.ip[7],
key.ip[8], key.ip[9], key.ip[10], key.ip[11],
key.ip[12], key.ip[13], key.ip[14], key.ip[15],
key.depth, i);
status = rte_pipeline_table_entry_add(p, table_id, &key, &entry,
&key_found, &entry_ptr);
if (status < 0)
rte_panic("Unable to add entry to table %u (%d)\n",
table_id, status);
}
/* Enable input ports */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_enable(p, port_in_id[i]))
rte_panic("Unable to enable input port %u\n",
port_in_id[i]);
/* Check pipeline consistency */
if (rte_pipeline_check(p) < 0)
rte_panic("Pipeline consistency check failed\n");
/* Run-time */
#if APP_FLUSH == 0
for ( ; ; )
rte_pipeline_run(p);
#else
for (i = 0; ; i++) {
rte_pipeline_run(p);
if ((i & APP_FLUSH) == 0)
rte_pipeline_flush(p);
}
#endif
}
static int
test_distributor(void)
{
static struct rte_distributor *ds;
static struct rte_distributor *db;
static struct rte_distributor *dist[2];
static struct rte_mempool *p;
int i;
if (rte_lcore_count() < 2) {
printf("ERROR: not enough cores to test distributor\n");
return -1;
}
if (db == NULL) {
db = rte_distributor_create("Test_dist_burst", rte_socket_id(),
rte_lcore_count() - 1,
RTE_DIST_ALG_BURST);
if (db == NULL) {
printf("Error creating burst distributor\n");
return -1;
}
} else {
rte_distributor_flush(db);
rte_distributor_clear_returns(db);
}
if (ds == NULL) {
ds = rte_distributor_create("Test_dist_single",
rte_socket_id(),
rte_lcore_count() - 1,
RTE_DIST_ALG_SINGLE);
if (ds == NULL) {
printf("Error creating single distributor\n");
return -1;
}
} else {
rte_distributor_flush(ds);
rte_distributor_clear_returns(ds);
}
const unsigned nb_bufs = (511 * rte_lcore_count()) < BIG_BATCH ?
(BIG_BATCH * 2) - 1 : (511 * rte_lcore_count());
if (p == NULL) {
p = rte_pktmbuf_pool_create("DT_MBUF_POOL", nb_bufs, BURST,
0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
if (p == NULL) {
printf("Error creating mempool\n");
return -1;
}
}
dist[0] = ds;
dist[1] = db;
for (i = 0; i < 2; i++) {
worker_params.dist = dist[i];
if (i)
sprintf(worker_params.name, "burst");
else
sprintf(worker_params.name, "single");
rte_eal_mp_remote_launch(handle_work,
&worker_params, SKIP_MASTER);
if (sanity_test(&worker_params, p) < 0)
goto err;
quit_workers(&worker_params, p);
rte_eal_mp_remote_launch(handle_work_with_free_mbufs,
&worker_params, SKIP_MASTER);
if (sanity_test_with_mbuf_alloc(&worker_params, p) < 0)
goto err;
quit_workers(&worker_params, p);
if (rte_lcore_count() > 2) {
rte_eal_mp_remote_launch(handle_work_for_shutdown_test,
&worker_params,
SKIP_MASTER);
if (sanity_test_with_worker_shutdown(&worker_params,
p) < 0)
goto err;
quit_workers(&worker_params, p);
rte_eal_mp_remote_launch(handle_work_for_shutdown_test,
&worker_params,
SKIP_MASTER);
if (test_flush_with_worker_shutdown(&worker_params,
p) < 0)
goto err;
quit_workers(&worker_params, p);
} else {
printf("Too few cores to run worker shutdown test\n");
}
}
if (test_error_distributor_create_numworkers() == -1 ||
test_error_distributor_create_name() == -1) {
printf("rte_distributor_create parameter check tests failed");
return -1;
}
return 0;
err:
quit_workers(&worker_params, p);
return -1;
}
static int
paxos_rx_process(struct rte_mbuf *pkt, struct proposer* proposer)
{
int ret = 0;
uint8_t l4_proto = 0;
uint16_t outer_header_len;
union tunnel_offload_info info = { .data = 0 };
struct udp_hdr *udp_hdr;
struct paxos_hdr *paxos_hdr;
struct ether_hdr *phdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
parse_ethernet(phdr, &info, &l4_proto);
if (l4_proto != IPPROTO_UDP)
return -1;
udp_hdr = (struct udp_hdr *)((char *)phdr +
info.outer_l2_len + info.outer_l3_len);
/* if UDP dst port is not either PROPOSER or LEARNER port */
if (!(udp_hdr->dst_port == rte_cpu_to_be_16(PROPOSER_PORT) ||
udp_hdr->dst_port == rte_cpu_to_be_16(LEARNER_PORT)) &&
(pkt->packet_type & RTE_PTYPE_TUNNEL_MASK) == 0)
return -1;
paxos_hdr = (struct paxos_hdr *)((char *)udp_hdr + sizeof(struct udp_hdr));
if (rte_get_log_level() == RTE_LOG_DEBUG) {
//rte_hexdump(stdout, "udp", udp_hdr, sizeof(struct udp_hdr));
//rte_hexdump(stdout, "paxos", paxos_hdr, sizeof(struct paxos_hdr));
print_paxos_hdr(paxos_hdr);
}
int value_len = rte_be_to_cpu_16(paxos_hdr->value_len);
struct paxos_value *v = paxos_value_new((char *)paxos_hdr->paxosval, value_len);
switch(rte_be_to_cpu_16(paxos_hdr->msgtype)) {
case PAXOS_PROMISE: {
struct paxos_promise promise = {
.iid = rte_be_to_cpu_32(paxos_hdr->inst),
.ballot = rte_be_to_cpu_16(paxos_hdr->rnd),
.value_ballot = rte_be_to_cpu_16(paxos_hdr->vrnd),
.aid = rte_be_to_cpu_16(paxos_hdr->acptid),
.value = *v
};
proposer_handle_promise(proposer, &promise);
break;
}
case PAXOS_ACCEPT: {
if (first_time) {
proposer_preexecute(proposer);
first_time = false;
}
struct paxos_accept acpt = {
.iid = rte_be_to_cpu_32(paxos_hdr->inst),
.ballot = rte_be_to_cpu_16(paxos_hdr->rnd),
.value_ballot = rte_be_to_cpu_16(paxos_hdr->vrnd),
.aid = rte_be_to_cpu_16(paxos_hdr->acptid),
.value = *v
};
proposer_handle_accept(proposer, &acpt);
break;
}
case PAXOS_ACCEPTED: {
struct paxos_accepted ack = {
.iid = rte_be_to_cpu_32(paxos_hdr->inst),
.ballot = rte_be_to_cpu_16(paxos_hdr->rnd),
.value_ballot = rte_be_to_cpu_16(paxos_hdr->vrnd),
.aid = rte_be_to_cpu_16(paxos_hdr->acptid),
.value = *v
};
proposer_handle_accepted(proposer, &ack);
break;
}
default:
break;
}
outer_header_len = info.outer_l2_len + info.outer_l3_len
+ sizeof(struct udp_hdr) + sizeof(struct paxos_hdr);
rte_pktmbuf_adj(pkt, outer_header_len);
return ret;
}
static uint16_t
add_timestamps(uint8_t port __rte_unused, uint16_t qidx __rte_unused,
struct rte_mbuf **pkts, uint16_t nb_pkts,
uint16_t max_pkts __rte_unused, void *user_param)
{
struct proposer* proposer = (struct proposer *)user_param;
unsigned i;
uint64_t now = rte_rdtsc();
for (i = 0; i < nb_pkts; i++) {
pkts[i]->udata64 = now;
paxos_rx_process(pkts[i], proposer);
}
return nb_pkts;
}
static inline int
port_init(uint8_t port, struct rte_mempool *mbuf_pool, struct proposer* proposer)
{
struct rte_eth_dev_info dev_info;
struct rte_eth_txconf *txconf;
struct rte_eth_rxconf *rxconf;
struct rte_eth_conf port_conf = port_conf_default;
const uint16_t rx_rings = 1, tx_rings = 1;
int retval;
uint16_t q;
rte_eth_dev_info_get(port, &dev_info);
rxconf = &dev_info.default_rxconf;
txconf = &dev_info.default_txconf;
txconf->txq_flags &= PKT_TX_IPV4;
txconf->txq_flags &= PKT_TX_UDP_CKSUM;
if (port >= rte_eth_dev_count())
return -1;
retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
if (retval != 0)
return retval;
for (q = 0; q < rx_rings; q++) {
retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE,
rte_eth_dev_socket_id(port), rxconf, mbuf_pool);
if (retval < 0)
return retval;
}
for (q = 0; q < tx_rings; q++) {
retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE,
rte_eth_dev_socket_id(port), txconf);
if (retval < 0)
return retval;
}
retval = rte_eth_dev_start(port);
if (retval < 0)
return retval;
struct ether_addr addr;
rte_eth_macaddr_get(port, &addr);
rte_eth_promiscuous_enable(port);
rte_eth_add_rx_callback(port, 0, add_timestamps, proposer);
rte_eth_add_tx_callback(port, 0, calc_latency, NULL);
return 0;
}
static void
lcore_main(uint8_t port, __rte_unused struct proposer *p)
{
proposer_preexecute(p);
for (;;) {
// Check if signal is received
if (force_quit)
break;
struct rte_mbuf *bufs[BURST_SIZE];
const uint16_t nb_rx = rte_eth_rx_burst(port, 0, bufs, BURST_SIZE);
if (unlikely(nb_rx == 0))
continue;
uint16_t buf;
for (buf = 0; buf < nb_rx; buf++)
rte_pktmbuf_free(bufs[buf]);
}
}
static __attribute__((noreturn)) int
lcore_mainloop(__attribute__((unused)) void *arg)
{
uint64_t prev_tsc = 0, cur_tsc, diff_tsc;
unsigned lcore_id;
lcore_id = rte_lcore_id();
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_TIMER,
"Starting mainloop on core %u\n", lcore_id);
while(1) {
cur_tsc = rte_rdtsc();
diff_tsc = cur_tsc - prev_tsc;
if (diff_tsc > TIMER_RESOLUTION_CYCLES) {
rte_timer_manage();
prev_tsc = cur_tsc;
}
}
}
static void
report_stat(struct rte_timer *tim, __attribute((unused)) void *arg)
{
/* print stat */
uint32_t count = rte_atomic32_read(&stat);
rte_log(RTE_LOG_INFO, RTE_LOGTYPE_USER8,
"Throughput = %8u msg/s\n", count);
/* reset stat */
rte_atomic32_set(&stat, 0);
/* this timer is automatically reloaded until we decide to stop it */
if (force_quit)
rte_timer_stop(tim);
}
static void
check_timeout(struct rte_timer *tim, void *arg)
{
struct proposer* p = (struct proposer *) arg;
unsigned lcore_id = rte_lcore_id();
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8, "%s() on lcore_id %i\n", __func__, lcore_id);
struct paxos_message out;
out.type = PAXOS_PREPARE;
struct timeout_iterator* iter = proposer_timeout_iterator(p);
while(timeout_iterator_prepare(iter, &out.u.prepare)) {
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8,
"%s Send PREPARE inst %d ballot %d\n",
__func__, out.u.prepare.iid, out.u.prepare.ballot);
send_paxos_message(&out);
}
out.type = PAXOS_ACCEPT;
while(timeout_iterator_accept(iter, &out.u.accept)) {
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8,
"%s: Send ACCEPT inst %d ballot %d\n",
__func__, out.u.prepare.iid, out.u.prepare.ballot);
send_paxos_message(&out);
}
timeout_iterator_free(iter);
/* this timer is automatically reloaded until we decide to stop it */
if (force_quit)
rte_timer_stop(tim);
}
int
main(int argc, char *argv[])
{
uint8_t portid = 0;
unsigned master_core, lcore_id;
signal(SIGTERM, signal_handler);
signal(SIGINT, signal_handler);
force_quit = false;
int proposer_id = 0;
if (rte_get_log_level() == RTE_LOG_DEBUG) {
paxos_config.verbosity = PAXOS_LOG_DEBUG;
}
struct proposer *proposer = proposer_new(proposer_id, NUM_ACCEPTORS);
first_time = true;
/* init EAL */
int ret = rte_eal_init(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
/* init timer structure */
rte_timer_init(&timer);
rte_timer_init(&stat_timer);
/* load deliver_timer, every 1 s, on a slave lcore, reloaded automatically */
uint64_t hz = rte_get_timer_hz();
/* Call rte_timer_manage every 10ms */
TIMER_RESOLUTION_CYCLES = hz / 100;
rte_log(RTE_LOG_INFO, RTE_LOGTYPE_USER1, "Clock: %"PRIu64"\n", hz);
/* master core */
master_core = rte_lcore_id();
/* slave core */
lcore_id = rte_get_next_lcore(master_core, 0, 1);
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER1, "lcore_id: %d\n", lcore_id);
rte_timer_reset(&timer, hz, PERIODICAL, lcore_id, check_timeout, proposer);
/* reset timer */
rte_eal_remote_launch(lcore_mainloop, NULL, lcore_id);
/* stat core */
lcore_id = rte_get_next_lcore(lcore_id , 0, 1);
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER1, "lcore_id: %d\n", lcore_id);
rte_timer_reset(&stat_timer, hz, PERIODICAL, lcore_id,
report_stat, NULL);
/* init RTE timer library */
rte_timer_subsystem_init();
mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL",
NUM_MBUFS, MBUF_CACHE_SIZE, 0,
RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
if (mbuf_pool == NULL)
rte_exit(EXIT_FAILURE, "Cannot create mbuf_pool\n");
/* reset timer */
rte_eal_remote_launch(lcore_mainloop, NULL, lcore_id);
if (port_init(portid, mbuf_pool, proposer) != 0)
rte_exit(EXIT_FAILURE, "Cannot init port %"PRIu8"\n", portid);
lcore_main(portid, proposer);
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8, "Free proposer\n");
proposer_free(proposer);
return 0;
}
void
app_main_loop_worker_pipeline_acl(void) {
struct rte_pipeline_params pipeline_params = {
.name = "pipeline",
.socket_id = rte_socket_id(),
};
struct rte_pipeline *p;
uint32_t port_in_id[APP_MAX_PORTS];
uint32_t port_out_id[APP_MAX_PORTS];
uint32_t table_id;
uint32_t i;
RTE_LOG(INFO, USER1,
"Core %u is doing work (pipeline with ACL table)\n",
rte_lcore_id());
/* Pipeline configuration */
p = rte_pipeline_create(&pipeline_params);
if (p == NULL)
rte_panic("Unable to configure the pipeline\n");
/* Input port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_reader_params port_ring_params = {
.ring = app.rings_rx[i],
};
struct rte_pipeline_port_in_params port_params = {
.ops = &rte_port_ring_reader_ops,
.arg_create = (void *) &port_ring_params,
.f_action = NULL,
.arg_ah = NULL,
.burst_size = app.burst_size_worker_read,
};
if (rte_pipeline_port_in_create(p, &port_params,
&port_in_id[i]))
rte_panic("Unable to configure input port for "
"ring %d\n", i);
}
/* Output port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_writer_params port_ring_params = {
.ring = app.rings_tx[i],
.tx_burst_sz = app.burst_size_worker_write,
};
struct rte_pipeline_port_out_params port_params = {
.ops = &rte_port_ring_writer_ops,
.arg_create = (void *) &port_ring_params,
.f_action = NULL,
.arg_ah = NULL,
};
if (rte_pipeline_port_out_create(p, &port_params,
&port_out_id[i]))
rte_panic("Unable to configure output port for "
"ring %d\n", i);
}
/* Table configuration */
{
struct rte_table_acl_params table_acl_params = {
.name = "test", /* unique identifier for acl contexts */
.n_rules = 1 << 5,
.n_rule_fields = DIM(ipv4_field_formats),
};
/* Copy in the rule meta-data defined above into the params */
memcpy(table_acl_params.field_format, ipv4_field_formats,
sizeof(ipv4_field_formats));
struct rte_pipeline_table_params table_params = {
.ops = &rte_table_acl_ops,
.arg_create = &table_acl_params,
.f_action_hit = NULL,
.f_action_miss = NULL,
.arg_ah = NULL,
.action_data_size = 0,
};
if (rte_pipeline_table_create(p, &table_params, &table_id))
rte_panic("Unable to configure the ACL table\n");
}
/* Interconnecting ports and tables */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_connect_to_table(p, port_in_id[i],
table_id))
rte_panic("Unable to connect input port %u to "
"table %u\n", port_in_id[i], table_id);
/* Add entries to tables */
for (i = 0; i < app.n_ports; i++) {
struct rte_pipeline_table_entry table_entry = {
.action = RTE_PIPELINE_ACTION_PORT,
{.port_id = port_out_id[i & (app.n_ports - 1)]},
};
struct rte_table_acl_rule_add_params rule_params;
struct rte_pipeline_table_entry *entry_ptr;
int key_found, ret;
memset(&rule_params, 0, sizeof(rule_params));
/* Set the rule values */
rule_params.field_value[SRC_FIELD_IPV4].value.u32 = 0;
rule_params.field_value[SRC_FIELD_IPV4].mask_range.u32 = 0;
rule_params.field_value[DST_FIELD_IPV4].value.u32 =
i << (24 - __builtin_popcount(app.n_ports - 1));
rule_params.field_value[DST_FIELD_IPV4].mask_range.u32 =
8 + __builtin_popcount(app.n_ports - 1);
rule_params.field_value[SRCP_FIELD_IPV4].value.u16 = 0;
rule_params.field_value[SRCP_FIELD_IPV4].mask_range.u16 =
UINT16_MAX;
rule_params.field_value[DSTP_FIELD_IPV4].value.u16 = 0;
rule_params.field_value[DSTP_FIELD_IPV4].mask_range.u16 =
UINT16_MAX;
rule_params.field_value[PROTO_FIELD_IPV4].value.u8 = 0;
rule_params.field_value[PROTO_FIELD_IPV4].mask_range.u8 = 0;
rule_params.priority = 0;
uint32_t dst_addr = rule_params.field_value[DST_FIELD_IPV4].
value.u32;
uint32_t dst_mask =
rule_params.field_value[DST_FIELD_IPV4].mask_range.u32;
printf("Adding rule to ACL table (IPv4 destination = "
"%u.%u.%u.%u/%u => port out = %u)\n",
(dst_addr & 0xFF000000) >> 24,
(dst_addr & 0x00FF0000) >> 16,
(dst_addr & 0x0000FF00) >> 8,
dst_addr & 0x000000FF,
dst_mask,
table_entry.port_id);
/* For ACL, add needs an rte_table_acl_rule_add_params struct */
ret = rte_pipeline_table_entry_add(p, table_id, &rule_params,
&table_entry, &key_found, &entry_ptr);
if (ret < 0)
rte_panic("Unable to add entry to table %u (%d)\n",
table_id, ret);
}
/* Enable input ports */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_enable(p, port_in_id[i]))
rte_panic("Unable to enable input port %u\n",
port_in_id[i]);
/* Check pipeline consistency */
if (rte_pipeline_check(p) < 0)
rte_panic("Pipeline consistency check failed\n");
/* Run-time */
#if APP_FLUSH == 0
for ( ; ; )
rte_pipeline_run(p);
#else
for (i = 0; ; i++) {
rte_pipeline_run(p);
if ((i & APP_FLUSH) == 0)
rte_pipeline_flush(p);
}
#endif
}
/**
* Set up the DPDK rings which will be used to pass packets, via
* pointers, between the multi-process server and client processes.
* Each client needs one RX queue.
*/
static int
init_shm_rings(void)
{
unsigned i;
unsigned socket_id;
const char * q_name;
#ifdef INTERRUPT_FIFO
const char * fifo_name;
#endif
#ifdef INTERRUPT_SEM
const char * sem_name;
sem_t *mutex;
#endif
#if defined(INTERRUPT_FIFO) || defined(INTERRUPT_SEM)
#ifdef DPDK_FLAG
const char *irq_flag_name;
const unsigned flagsize = 4;
#else
key_t key;
int shmid;
char *shm;
#endif
#endif
const unsigned ringsize = CLIENT_QUEUE_RINGSIZE;
clients = rte_malloc("client details",
sizeof(*clients) * num_clients, 0);
if (clients == NULL)
rte_exit(EXIT_FAILURE, "Cannot allocate memory for client program details\n");
for (i = 0; i < num_clients; i++) {
/* Create an RX queue for each client */
socket_id = rte_socket_id();
q_name = get_rx_queue_name(i);
clients[i].rx_q = rte_ring_create(q_name,
ringsize, socket_id,
RING_F_SP_ENQ | RING_F_SC_DEQ ); /* single prod, single cons */
//verify two functions
uint16_t ring_cur_entries = rte_ring_count(clients[i].rx_q);
uint16_t ring_free_entries = rte_ring_free_count(clients[i].rx_q);
fprintf(stderr, "ring_cur_entries=%d, ring_free_entries=%d\n", ring_cur_entries, ring_free_entries);
if (clients[i].rx_q == NULL)
rte_exit(EXIT_FAILURE, "Cannot create rx ring queue for client %u\n", i);
//add by wei, create FIFO pipe
#ifdef INTERRUPT_FIFO
umask(0);
fifo_name = get_fifo_name(i);
clients[i].fifo_name = fifo_name;
mknod(fifo_name, S_IFIFO|0666, 0);
clients[i].fifo_fp = fopen(fifo_name, "w");
if(clients[i].fifo_fp == NULL) {
fprintf(stderr, "can not create FIFO for client %d\n", i);
exit(1);
}
#endif
#ifdef INTERRUPT_SEM
sem_name = get_sem_name(i);
clients[i].sem_name = sem_name;
fprintf(stderr, "sem_name=%s for client %d\n", sem_name, i);
mutex = sem_open(sem_name, O_CREAT, 06666, 0);
if(mutex == SEM_FAILED) {
fprintf(stderr, "can not create semaphore for client %d\n", i);
sem_unlink(sem_name);
exit(1);
}
clients[i].mutex = mutex;
#endif
#if defined(INTERRUPT_FIFO) || defined(INTERRUPT_SEM)
#ifdef DPDK_FLAG
irq_flag_name = get_irq_flag_name(i);
clients[i].irq_flag = (int *)rte_ring_create(irq_flag_name,
flagsize, socket_id,
RING_F_SP_ENQ | RING_F_SC_DEQ ); /* single prod, single cons */
#else
key = get_rx_shmkey(i);
if ((shmid = shmget(key, SHMSZ, IPC_CREAT | 0666)) < 0) {
fprintf(stderr, "can not create the shared memory segment for client %d\n", i);
exit(1);
}
if ((shm = shmat(shmid, NULL, 0)) == (char *) -1) {
fprintf(stderr, "can not attach the shared segment to the server space for client %d\n", i);
exit(1);
}
clients[i].shm_server = (int *)shm;
#endif
#endif
}
return 0;
}
int dpdk_init(void)
{
int ret;
/* -m stands for memory in MBs that DPDK will allocate. Must be enough
* to accommodate the pool_size defined below. */
char *argv[] = { "./ix", "-m", "148" };
const int pool_buffer_size = 0;
const int pool_cache_size = 0;
/* pool_size sets an implicit limit on cores * NICs that DPDK allows */
const int pool_size = 32768;
optind = 0;
internal_config.no_hugetlbfs = 1;
ret = rte_eal_init(sizeof(argv) / sizeof(argv[0]), argv);
if (ret < 0)
return ret;
dpdk_pool = rte_pktmbuf_pool_create("mempool", pool_size, pool_cache_size, 0, pool_buffer_size, rte_socket_id());
if (dpdk_pool == NULL)
panic("Cannot create DPDK pool\n");
return 0;
}
/**
* Main init function for the multi-process server app,
* calls subfunctions to do each stage of the initialisation.
*/
int
init(int argc, char *argv[])
{
int retval;
const struct rte_memzone *mz;
uint8_t i, total_ports;
/* init EAL, parsing EAL args */
retval = rte_eal_init(argc, argv);
if (retval < 0)
return -1;
argc -= retval;
argv += retval;
/* get total number of ports */
total_ports = rte_eth_dev_count();
/* set up array for port data */
if (rte_eal_process_type() == RTE_PROC_SECONDARY)
{
mz = rte_memzone_lookup(VM_MZ_PORT_INFO);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot get port info structure\n");
ports = mz->addr;
}
else
{
mz = rte_memzone_reserve(VM_MZ_PORT_INFO, sizeof(*ports),
rte_socket_id(), NO_FLAGS);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot reserve memory zone for port information\n");
memset(mz->addr, 0, sizeof(*ports));
ports = mz->addr;
}
/* parse additional, application arguments */
retval = parse_app_args(total_ports, argc, argv);
if (retval != 0)
return -1;
/* initialise mbuf pools */
retval = init_mbuf_pools();
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot create needed mbuf pools\n");
/* now initialise the ports we will use */
if (rte_eal_process_type() == RTE_PROC_SECONDARY)
{
;
}
else
{
for (i = 0; i < ports->num_ports; i++) {
retval = init_port(ports->id[i]);
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot initialise port %u\n",
(unsigned)i);
}
}
check_all_ports_link_status(ports->num_ports, (~0x0));
/* initialise the client queues/rings for inter-eu comms
init_shm_rings();
*/
return 0;
}
/**
* Main init function for the multi-process server app,
* calls subfunctions to do each stage of the initialisation.
*/
int
init(int argc, char *argv[])
{
int retval;
const struct rte_memzone *mz;
unsigned i, total_ports;
/* init EAL, parsing EAL args */
retval = rte_eal_init(argc, argv);
if (retval < 0)
return -1;
argc -= retval;
argv += retval;
/* get total number of ports */
total_ports = rte_eth_dev_count();
/* set up array for port data */
if (rte_eal_process_type() == RTE_PROC_SECONDARY)
{
mz = rte_memzone_lookup(MZ_PORT_INFO);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot get port info structure\n");
ports = mz->addr;
}
else /* RTE_PROC_PRIMARY */
{
mz = rte_memzone_reserve(MZ_PORT_INFO, sizeof(*ports),
rte_socket_id(), NO_FLAGS);
if (mz == NULL)
rte_exit(EXIT_FAILURE, "Cannot reserve memory zone for port information\n");
memset(mz->addr, 0, sizeof(*ports));
ports = mz->addr;
}
/* parse additional, application arguments */
retval = parse_app_args(total_ports, argc, argv);
if (retval != 0)
return -1;
/* initialise mbuf pools */
retval = init_mbuf_pools();
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot create needed mbuf pools\n");
/* now initialise the ports we will use */
if (rte_eal_process_type() == RTE_PROC_PRIMARY)
{
for (i = 0; i < ports->num_ports; i++) {
retval = init_port(ports->id[i]);
if (retval != 0)
rte_exit(EXIT_FAILURE, "Cannot initialise port %u\n",
(unsigned)i);
}
}
check_all_ports_link_status(ports->num_ports, (~0x0));
/* initialise the client queues/rings for inter-eu comms */
init_shm_rings();
if (rte_eal_process_type() == RTE_PROC_PRIMARY)
{
RTE_LOG(INFO, APP, "HOST SHARE MEM Init.\n");
/* create metadata, output cmdline
if (rte_hostshmem_metadata_create(HOSTSHMEM_METADATA_NAME) < 0)
rte_exit(EXIT_FAILURE, "Cannot create HOSTSHMEM metadata\n");
if (rte_hostshmem_metadata_add_memzone(mz, HOSTSHMEM_METADATA_NAME))
rte_exit(EXIT_FAILURE, "Cannot add memzone to HOSTSHMEM metadata\n");
if (rte_hostshmem_metadata_add_mempool(pktmbuf_pool, HOSTSHMEM_METADATA_NAME))
rte_exit(EXIT_FAILURE, "Cannot add mbuf mempool to HOSTSHMEM metadata\n");
for (i = 0; i < num_clients; i++)
{
if (rte_hostshmem_metadata_add_ring(clients[i].rx_q,
HOSTSHMEM_METADATA_NAME) < 0)
rte_exit(EXIT_FAILURE, "Cannot add ring client %d to HOSTSHMEM metadata\n", i);
}
generate_hostshmem_cmdline(HOSTSHMEM_METADATA_NAME);
*/
const struct rte_mem_config *mcfg;
/* get pointer to global configuration */
mcfg = rte_eal_get_configuration()->mem_config;
for (i = 0; i < RTE_MAX_MEMSEG; i++) {
if (mcfg->memseg[i].addr == NULL)
break;
printf("Segment %u: phys:0x%"PRIx64", len:%zu, "
"virt:%p, socket_id:%"PRId32", "
"hugepage_sz:%"PRIu64", nchannel:%"PRIx32", "
"nrank:%"PRIx32"\n", i,
mcfg->memseg[i].phys_addr,
mcfg->memseg[i].len,
mcfg->memseg[i].addr,
mcfg->memseg[i].socket_id,
mcfg->memseg[i].hugepage_sz,
mcfg->memseg[i].nchannel,
mcfg->memseg[i].nrank);
}
RTE_LOG(INFO, APP, "HOST SHARE MEM Init. done\n");
RTE_LOG(INFO, APP, "IV SHARE MEM Init.\n");
/* create metadata, output cmdline */
if (rte_ivshmem_metadata_create(IVSHMEM_METADATA_NAME) < 0)
rte_exit(EXIT_FAILURE, "Cannot create IVSHMEM metadata\n");
if (rte_ivshmem_metadata_add_memzone(mz, IVSHMEM_METADATA_NAME))
rte_exit(EXIT_FAILURE, "Cannot add memzone to IVSHMEM metadata\n");
if (rte_ivshmem_metadata_add_mempool(pktmbuf_pool, IVSHMEM_METADATA_NAME))
rte_exit(EXIT_FAILURE, "Cannot add mbuf mempool to IVSHMEM metadata\n");
for (i = 0; i < num_clients; i++)
{
if (rte_ivshmem_metadata_add_ring(clients[i].rx_q,
IVSHMEM_METADATA_NAME) < 0)
rte_exit(EXIT_FAILURE, "Cannot add ring client %d to IVSHMEM metadata\n", i);
}
generate_ivshmem_cmdline(IVSHMEM_METADATA_NAME);
RTE_LOG(INFO, APP, "IV SHARE MEM Done.\n");
}
return 0;
}
/* Main function */
int main(int argc, char **argv)
{
int ret;
int i;
/* Create handler for SIGINT for CTRL + C closing and SIGALRM to print stats*/
signal(SIGINT, sig_handler);
signal(SIGALRM, alarm_routine);
/* Initialize DPDK enviroment with args, then shift argc and argv to get application parameters */
ret = rte_eal_init(argc, argv);
if (ret < 0) FATAL_ERROR("Cannot init EAL\n");
argc -= ret;
argv += ret;
/* Check if this application can use 1 core*/
ret = rte_lcore_count ();
if (ret != 1) FATAL_ERROR("This application needs exactly 1 cores.");
/* Parse arguments */
parse_args(argc, argv);
if (ret < 0) FATAL_ERROR("Wrong arguments\n");
/* Probe PCI bus for ethernet devices, mandatory only in DPDK < 1.8.0 */
#if RTE_VER_MAJOR == 1 && RTE_VER_MINOR < 8
ret = rte_eal_pci_probe();
if (ret < 0) FATAL_ERROR("Cannot probe PCI\n");
#endif
/* Get number of ethernet devices */
nb_sys_ports = rte_eth_dev_count();
if (nb_sys_ports <= 0) FATAL_ERROR("Cannot find ETH devices\n");
/* Create a mempool with per-core cache, initializing every element for be used as mbuf, and allocating on the current NUMA node */
pktmbuf_pool = rte_mempool_create(MEMPOOL_NAME, buffer_size-1, MEMPOOL_ELEM_SZ, MEMPOOL_CACHE_SZ, sizeof(struct rte_pktmbuf_pool_private), rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,rte_socket_id(), 0);
if (pktmbuf_pool == NULL) FATAL_ERROR("Cannot create cluster_mem_pool. Errno: %d [ENOMEM: %d, ENOSPC: %d, E_RTE_NO_TAILQ: %d, E_RTE_NO_CONFIG: %d, E_RTE_SECONDARY: %d, EINVAL: %d, EEXIST: %d]\n", rte_errno, ENOMEM, ENOSPC, E_RTE_NO_TAILQ, E_RTE_NO_CONFIG, E_RTE_SECONDARY, EINVAL, EEXIST );
/* Operations needed for each ethernet device */
for(i=0; i < nb_sys_ports; i++)
init_port(i);
/* ... and then loop in consumer */
main_loop_producer ( NULL );
return 0;
}
/* Main function, does initialisation and calls the per-lcore functions */
int
main(int argc, char *argv[])
{
unsigned cores;
struct rte_mempool *mbuf_pool;
unsigned lcore_id;
uintptr_t i;
int ret;
unsigned nb_ports, valid_num_ports;
uint16_t portid;
signal(SIGHUP, sighup_handler);
/* init EAL */
ret = rte_eal_init(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
argc -= ret;
argv += ret;
/* parse app arguments */
ret = vmdq_parse_args(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Invalid VMDQ argument\n");
cores = rte_lcore_count();
if ((cores & (cores - 1)) != 0 || cores > RTE_MAX_LCORE) {
rte_exit(EXIT_FAILURE,"This program can only run on an even"
" number of cores(1-%d)\n\n", RTE_MAX_LCORE);
}
nb_ports = rte_eth_dev_count();
/*
* Update the global var NUM_PORTS and global array PORTS
* and get value of var VALID_NUM_PORTS according to system ports number
*/
valid_num_ports = check_ports_num(nb_ports);
if (valid_num_ports < 2 || valid_num_ports % 2) {
printf("Current valid ports number is %u\n", valid_num_ports);
rte_exit(EXIT_FAILURE, "Error with valid ports number is not even or less than 2\n");
}
mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL",
NUM_MBUFS_PER_PORT * nb_ports, MBUF_CACHE_SIZE,
0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
if (mbuf_pool == NULL)
rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
/* initialize all ports */
for (portid = 0; portid < nb_ports; portid++) {
/* skip ports that are not enabled */
if ((enabled_port_mask & (1 << portid)) == 0) {
printf("\nSkipping disabled port %d\n", portid);
continue;
}
if (port_init(portid, mbuf_pool) != 0)
rte_exit(EXIT_FAILURE, "Cannot initialize network ports\n");
}
/* call lcore_main() on every slave lcore */
i = 0;
RTE_LCORE_FOREACH_SLAVE(lcore_id) {
rte_eal_remote_launch(lcore_main, (void*)i++, lcore_id);
}
/* call on master too */
(void) lcore_main((void*)i);
return 0;
}
void
app_main_loop_pipeline_tx(void) {
struct rte_pipeline *p;
uint32_t port_in_id[APP_MAX_PORTS];
uint32_t port_out_id[APP_MAX_PORTS];
uint32_t table_id[APP_MAX_PORTS];
uint32_t i;
uint32_t core_id = rte_lcore_id();
struct app_core_params *core_params = app_get_core_params(core_id);
if ((core_params == NULL) || (core_params->core_type != APP_CORE_TX))
rte_panic("Core %u misconfiguration\n", core_id);
RTE_LOG(INFO, USER1, "Core %u is doing TX\n", core_id);
/* Pipeline configuration */
struct rte_pipeline_params pipeline_params = {
.name = "pipeline",
.socket_id = rte_socket_id(),
};
p = rte_pipeline_create(&pipeline_params);
if (p == NULL)
rte_panic("%s: Unable to configure the pipeline\n", __func__);
/* Input port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ring_reader_params port_ring_params = {
.ring = app.rings[core_params->swq_in[i]],
};
struct rte_pipeline_port_in_params port_params = {
.ops = &rte_port_ring_reader_ops,
.arg_create = (void *) &port_ring_params,
.f_action = (app.ether_hdr_pop_push) ?
app_pipeline_tx_port_in_action_handler : NULL,
.arg_ah = NULL,
.burst_size = app.bsz_swq_rd,
};
if (rte_pipeline_port_in_create(p, &port_params,
&port_in_id[i])) {
rte_panic("%s: Unable to configure input port for "
"ring TX %i\n", __func__, i);
}
}
/* Output port configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_port_ethdev_writer_params port_ethdev_params = {
.port_id = app.ports[i],
.queue_id = 0,
.tx_burst_sz = app.bsz_hwq_wr,
};
struct rte_pipeline_port_out_params port_params = {
.ops = &rte_port_ethdev_writer_ops,
.arg_create = (void *) &port_ethdev_params,
.f_action = NULL,
.f_action_bulk = NULL,
.arg_ah = NULL,
};
if (rte_pipeline_port_out_create(p, &port_params,
&port_out_id[i])) {
rte_panic("%s: Unable to configure output port for "
"port %d\n", __func__, app.ports[i]);
}
}
/* Table configuration */
for (i = 0; i < app.n_ports; i++) {
struct rte_pipeline_table_params table_params = {
.ops = &rte_table_stub_ops,
.arg_create = NULL,
.f_action_hit = NULL,
.f_action_miss = NULL,
.arg_ah = NULL,
.action_data_size = 0,
};
if (rte_pipeline_table_create(p, &table_params, &table_id[i])) {
rte_panic("%s: Unable to configure table %u\n",
__func__, table_id[i]);
}
}
/* Interconnecting ports and tables */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_connect_to_table(p, port_in_id[i],
table_id[i]))
rte_panic("%s: Unable to connect input port %u to "
"table %u\n", __func__, port_in_id[i],
table_id[i]);
/* Add entries to tables */
for (i = 0; i < app.n_ports; i++) {
struct rte_pipeline_table_entry default_entry = {
.action = RTE_PIPELINE_ACTION_PORT,
{.port_id = port_out_id[i]},
};
struct rte_pipeline_table_entry *default_entry_ptr;
if (rte_pipeline_table_default_entry_add(p, table_id[i],
&default_entry, &default_entry_ptr))
rte_panic("%s: Unable to add default entry to "
"table %u\n", __func__, table_id[i]);
}
/* Enable input ports */
for (i = 0; i < app.n_ports; i++)
if (rte_pipeline_port_in_enable(p, port_in_id[i]))
rte_panic("Unable to enable input port %u\n",
port_in_id[i]);
/* Check pipeline consistency */
if (rte_pipeline_check(p) < 0)
rte_panic("%s: Pipeline consistency check failed\n", __func__);
/* Run-time */
for (i = 0; ; i++) {
rte_pipeline_run(p);
if ((i & APP_FLUSH) == 0)
rte_pipeline_flush(p);
}
}
void
app_main_loop_tx(void) {
struct app_mbuf_array *m[APP_MAX_PORTS];
uint32_t i;
uint32_t core_id = rte_lcore_id();
struct app_core_params *core_params = app_get_core_params(core_id);
if ((core_params == NULL) || (core_params->core_type != APP_CORE_TX))
rte_panic("Core %u misconfiguration\n", core_id);
RTE_LOG(INFO, USER1, "Core %u is doing TX (no pipeline)\n", core_id);
for (i = 0; i < APP_MAX_PORTS; i++) {
m[i] = rte_malloc_socket(NULL, sizeof(struct app_mbuf_array),
CACHE_LINE_SIZE, rte_socket_id());
if (m[i] == NULL)
rte_panic("%s: Cannot allocate buffer space\n",
__func__);
}
for (i = 0; ; i = ((i + 1) & (app.n_ports - 1))) {
uint32_t n_mbufs, n_pkts;
int ret;
n_mbufs = m[i]->n_mbufs;
ret = rte_ring_sc_dequeue_bulk(
app.rings[core_params->swq_in[i]],
(void **) &m[i]->array[n_mbufs],
app.bsz_swq_rd);
if (ret == -ENOENT)
continue;
n_mbufs += app.bsz_swq_rd;
if (n_mbufs < app.bsz_hwq_wr) {
m[i]->n_mbufs = n_mbufs;
continue;
}
n_pkts = rte_eth_tx_burst(
app.ports[i],
0,
m[i]->array,
n_mbufs);
if (n_pkts < n_mbufs) {
uint32_t k;
for (k = n_pkts; k < n_mbufs; k++) {
struct rte_mbuf *pkt_to_free;
pkt_to_free = m[i]->array[k];
rte_pktmbuf_free(pkt_to_free);
}
}
m[i]->n_mbufs = 0;
}
}
void
pktgen_config_ports(void)
{
uint32_t lid, pid, i, s, q, sid;
rxtx_t rt;
pkt_seq_t *pkt;
port_info_t *info;
char buff[RTE_MEMZONE_NAMESIZE];
int32_t ret, cache_size;
char output_buff[256] = { 0 };
/* Find out the total number of ports in the system. */
/* We have already blacklisted the ones we needed to in main routine. */
pktgen.nb_ports = rte_eth_dev_count();
if (pktgen.nb_ports > RTE_MAX_ETHPORTS)
pktgen.nb_ports = RTE_MAX_ETHPORTS;
if (pktgen.nb_ports == 0)
pktgen_log_panic("*** Did not find any ports to use ***");
pktgen.starting_port = 0;
/* Setup the number of ports to display at a time */
if (pktgen.nb_ports > pktgen.nb_ports_per_page)
pktgen.ending_port = pktgen.starting_port + pktgen.nb_ports_per_page;
else
pktgen.ending_port = pktgen.starting_port + pktgen.nb_ports;
wr_port_matrix_dump(pktgen.l2p);
pktgen_log_info("Configuring %d ports, MBUF Size %d, MBUF Cache Size %d",
pktgen.nb_ports, MBUF_SIZE, MBUF_CACHE_SIZE);
/* For each lcore setup each port that is handled by that lcore. */
for (lid = 0; lid < RTE_MAX_LCORE; lid++) {
if (wr_get_map(pktgen.l2p, RTE_MAX_ETHPORTS, lid) == 0)
continue;
/* For each port attached or handled by the lcore */
for (pid = 0; pid < pktgen.nb_ports; pid++) {
/* If non-zero then this port is handled by this lcore. */
if (wr_get_map(pktgen.l2p, pid, lid) == 0)
continue;
wr_set_port_private(pktgen.l2p, pid, &pktgen.info[pid]);
pktgen.info[pid].pid = pid;
}
}
wr_dump_l2p(pktgen.l2p);
pktgen.total_mem_used = 0;
for (pid = 0; pid < pktgen.nb_ports; pid++) {
/* Skip if we do not have any lcores attached to a port. */
if ( (rt.rxtx = wr_get_map(pktgen.l2p, pid, RTE_MAX_LCORE)) == 0)
continue;
pktgen.port_cnt++;
snprintf(output_buff, sizeof(output_buff),
"Initialize Port %d -- TxQ %d, RxQ %d", pid, rt.tx, rt.rx);
info = wr_get_port_private(pktgen.l2p, pid);
info->fill_pattern_type = ABC_FILL_PATTERN;
strncpy(info->user_pattern, "0123456789abcdef", USER_PATTERN_SIZE);
rte_spinlock_init(&info->port_lock);
/* Create the pkt header structures for transmitting sequence of packets. */
snprintf(buff, sizeof(buff), "seq_hdr_%d", pid);
info->seq_pkt = (pkt_seq_t *)rte_zmalloc_socket(buff, (sizeof(pkt_seq_t) * NUM_TOTAL_PKTS),
RTE_CACHE_LINE_SIZE, rte_socket_id());
if (info->seq_pkt == NULL)
pktgen_log_panic("Unable to allocate %d pkt_seq_t headers", NUM_TOTAL_PKTS);
info->seqIdx = 0;
info->seqCnt = 0;
info->nb_mbufs = MAX_MBUFS_PER_PORT;
cache_size = (info->nb_mbufs > RTE_MEMPOOL_CACHE_MAX_SIZE) ?
RTE_MEMPOOL_CACHE_MAX_SIZE : info->nb_mbufs;
pktgen_port_conf_setup(pid, &rt, &default_port_conf);
if ( (ret = rte_eth_dev_configure(pid, rt.rx, rt.tx, &info->port_conf)) < 0)
pktgen_log_panic("Cannot configure device: port=%d, Num queues %d,%d (%d)%s",
pid, rt.rx, rt.tx, errno, rte_strerror(-ret));
pkt = &info->seq_pkt[SINGLE_PKT];
/* Grab the source MAC addresses * / */
rte_eth_macaddr_get(pid, &pkt->eth_src_addr);
pktgen_log_info("%s, Src MAC %02x:%02x:%02x:%02x:%02x:%02x", output_buff,
pkt->eth_src_addr.addr_bytes[0],
pkt->eth_src_addr.addr_bytes[1],
pkt->eth_src_addr.addr_bytes[2],
pkt->eth_src_addr.addr_bytes[3],
pkt->eth_src_addr.addr_bytes[4],
pkt->eth_src_addr.addr_bytes[5]);
/* Copy the first Src MAC address in SINGLE_PKT to the rest of the sequence packets. */
for (i = 0; i < NUM_SEQ_PKTS; i++)
ethAddrCopy(&info->seq_pkt[i].eth_src_addr, &pkt->eth_src_addr);
pktgen.mem_used = 0;
for (q = 0; q < rt.rx; q++) {
/* grab the socket id value based on the lcore being used. */
sid = rte_lcore_to_socket_id(wr_get_port_lid(pktgen.l2p, pid, q));
/* Create and initialize the default Receive buffers. */
info->q[q].rx_mp = pktgen_mbuf_pool_create("Default RX", pid, q, info->nb_mbufs, sid, cache_size);
if (info->q[q].rx_mp == NULL)
pktgen_log_panic("Cannot init port %d for Default RX mbufs", pid);
ret = rte_eth_rx_queue_setup(pid, q, pktgen.nb_rxd, sid, &info->rx_conf, pktgen.info[pid].q[q].rx_mp);
if (ret < 0)
pktgen_log_panic("rte_eth_rx_queue_setup: err=%d, port=%d, %s", ret, pid, rte_strerror(-ret));
}
pktgen_log_info("");
for (q = 0; q < rt.tx; q++) {
/* grab the socket id value based on the lcore being used. */
sid = rte_lcore_to_socket_id(wr_get_port_lid(pktgen.l2p, pid, q));
/* Create and initialize the default Transmit buffers. */
info->q[q].tx_mp = pktgen_mbuf_pool_create("Default TX", pid, q, MAX_MBUFS_PER_PORT, sid, cache_size);
if (info->q[q].tx_mp == NULL)
pktgen_log_panic("Cannot init port %d for Default TX mbufs", pid);
/* Create and initialize the range Transmit buffers. */
info->q[q].range_mp = pktgen_mbuf_pool_create("Range TX", pid, q, MAX_MBUFS_PER_PORT, sid, 0);
if (info->q[q].range_mp == NULL)
pktgen_log_panic("Cannot init port %d for Range TX mbufs", pid);
/* Create and initialize the sequence Transmit buffers. */
info->q[q].seq_mp = pktgen_mbuf_pool_create("Sequence TX", pid, q, MAX_MBUFS_PER_PORT, sid, cache_size);
if (info->q[q].seq_mp == NULL)
pktgen_log_panic("Cannot init port %d for Sequence TX mbufs", pid);
/* Used for sending special packets like ARP requests */
info->q[q].special_mp = pktgen_mbuf_pool_create("Special TX", pid, q, MAX_SPECIAL_MBUFS, sid, 0);
if (info->q[q].special_mp == NULL)
pktgen_log_panic("Cannot init port %d for Special TX mbufs", pid);
/* Setup the PCAP file for each port */
if (pktgen.info[pid].pcap != NULL)
if (pktgen_pcap_parse(pktgen.info[pid].pcap, info, q) == -1)
pktgen_log_panic("Cannot load PCAP file for port %d", pid);
/* Find out the link speed to program the WTHRESH value correctly. */
pktgen_get_link_status(info, pid, 0);
ret = rte_eth_tx_queue_setup(pid, q, pktgen.nb_txd, sid, &info->tx_conf);
if (ret < 0)
pktgen_log_panic("rte_eth_tx_queue_setup: err=%d, port=%d, %s", ret, pid, rte_strerror(-ret));
pktgen_log_info("");
}
pktgen_log_info("%*sPort memory used = %6lu KB", 71, " ", (pktgen.mem_used + 1023) / 1024);
}
pktgen_log_info("%*sTotal memory used = %6lu KB", 70, " ", (pktgen.total_mem_used + 1023) / 1024);
/* Start up the ports and display the port Link status */
for (pid = 0; pid < pktgen.nb_ports; pid++) {
if (wr_get_map(pktgen.l2p, pid, RTE_MAX_LCORE) == 0)
continue;
info = wr_get_port_private(pktgen.l2p, pid);
/* Start device */
if ( (ret = rte_eth_dev_start(pid)) < 0)
pktgen_log_panic("rte_eth_dev_start: port=%d, %s", pid, rte_strerror(-ret));
pktgen_get_link_status(info, pid, 1);
if (info->link.link_status)
snprintf(output_buff, sizeof(output_buff), "Port %2d: Link Up - speed %u Mbps - %s", pid,
(uint32_t)info->link.link_speed,
(info->link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
("full-duplex") : ("half-duplex"));
else
snprintf(output_buff, sizeof(output_buff), "Port %2d: Link Down", pid);
/* If enabled, put device in promiscuous mode. */
if (pktgen.flags & PROMISCUOUS_ON_FLAG) {
strncatf(output_buff, " <Enable promiscuous mode>");
rte_eth_promiscuous_enable(pid);
}
pktgen_log_info("%s", output_buff);
pktgen.info[pid].seq_pkt[SINGLE_PKT].pktSize = MIN_PKT_SIZE;
/* Setup the port and packet defaults. (must be after link speed is found) */
for (s = 0; s < NUM_TOTAL_PKTS; s++)
pktgen_port_defaults(pid, s);
pktgen_range_setup(info);
rte_eth_stats_get(pid, &info->init_stats);
pktgen_rnd_bits_init(&pktgen.info[pid].rnd_bitfields);
}
/* Clear the log information by putting a blank line */
pktgen_log_info("");
/* Setup the packet capture per port if needed. */
for (sid = 0; sid < wr_coremap_cnt(pktgen.core_info, pktgen.core_cnt, 0); sid++)
pktgen_packet_capture_init(&pktgen.capture[sid], sid);
}
int
main(int argc, char **argv)
{
//struct lcore_queue_conf *qconf = NULL;
//struct rte_eth_dev_info dev_info;
struct lcore_env** envs;
int ret;
uint8_t n_ports;
unsigned lcore_count;
ret = rte_eal_init(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Invalid EAL arguments\n");
argc -= ret;
argv += ret;
ret = l2sw_parse_args(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Invalid MARIO arguments\n");
lcore_count = rte_lcore_count();
n_ports = rte_eth_dev_count();
//RTE_LOG(INFO, MARIO, "Find %u logical cores\n" , lcore_count);
mbuf_pool = rte_mempool_create("mbuf_pool", NB_MBUF, MBUF_SIZE,
32, sizeof(struct rte_pktmbuf_pool_private),
rte_pktmbuf_pool_init, NULL,
rte_pktmbuf_init, NULL,
rte_socket_id(), 0);
if (mbuf_pool == NULL)
rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
// init route_table
route_table = create_route_table(ROUTE_ENTRY_SIZE);
add_staticroute(route_table);
// init arp_table
arp_table = create_arp_table(ARP_ENTRY_SIZE);
n_ports = rte_eth_dev_count();
if (n_ports == 0)
rte_exit(EXIT_FAILURE, "No Ethernet ports - byte\n");
//RTE_LOG(INFO, MARIO, "Find %u ethernet ports\n", n_ports);
if (n_ports > RTE_MAX_ETHPORTS)
n_ports = RTE_MAX_ETHPORTS;
/* Each logical core is assigned a dedicated TX queue on each port. */
/*
for(uint8_t port_id = 0; port_id < n_ports; port_id++) {
rte_eth_dev_info_get(port_id, &dev_info);
}
*/
/* Initialize the port/queue configuration of each logical core */
/*
for(uint8_t port_id = 0; port_id < n_ports; port_id++) {
;
}
*/
/* Initialize lcore_env */
envs = (struct lcore_env**) rte_malloc(NULL,sizeof(struct lcore_env*),0);
if (envs == NULL)
rte_exit(EXIT_FAILURE, "Cannot allocate memory for core envs\n");
uint8_t lcore_id;
for (lcore_id = 0; lcore_id < lcore_count; lcore_id++) {
struct lcore_env* env;
env = (struct lcore_env*) rte_malloc(NULL,sizeof(struct lcore_env) +
sizeof(struct mbuf_table) *n_ports,0);
if (env == NULL)
rte_exit(EXIT_FAILURE,
"Cannot allocate memory for %u core env\n", lcore_id);
env->n_port = n_ports;
env->lcore_id = lcore_id;
memset(env->tx_mbufs, 0, sizeof(struct mbuf_table) * n_ports);
envs[lcore_id] = env;
}
/* Initialise each port */
uint8_t port_id;
for(port_id = 0; port_id < n_ports; port_id++) {
//RTE_LOG(INFO, MARIO, "Initializing port %u...", port_id);
fflush(stdout);
ret = rte_eth_dev_configure(port_id, lcore_count, lcore_count, &port_conf);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%u\n",
ret, (unsigned)port_id);
//RTE_LOG(INFO, MARIO, "done\n");
rte_eth_macaddr_get(port_id, &port2eth[port_id]);
/* init one RX queue */
uint8_t core_id;
for (core_id = 0; core_id < lcore_count; core_id++) {
ret = rte_eth_rx_queue_setup(port_id, core_id, nb_rxd,
rte_eth_dev_socket_id(port_id),
NULL,
mbuf_pool);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"rte_eth_rx_queue_setup:err=%d, port=%u queue=%u\n",
ret, (unsigned) port_id, (unsigned) core_id);
}
/* init one TX queue */
for (core_id = 0; core_id < lcore_count; core_id++) {
ret = rte_eth_tx_queue_setup(port_id, core_id, nb_txd,
rte_eth_dev_socket_id(port_id), NULL);
if (ret < 0)
rte_exit(EXIT_FAILURE,
"rte_eth_tx_queue_setup:err=%d, port=%u queue=%u\n",
ret, (unsigned) port_id, (unsigned) core_id);
}
/* Start device */
ret = rte_eth_dev_start(port_id);
if (ret < 0)
rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err=%d, port=%u\n",
ret, (unsigned) port_id);
rte_eth_promiscuous_enable(port_id);
/*RTE_LOG(INFO, MARIO,
"Port %u, MAC address %02x:%02x:%02x:%02x:%02x:%02x\n\n",
port_id,
port2eth[port_id].addr_bytes[0],
port2eth[port_id].addr_bytes[1],
port2eth[port_id].addr_bytes[2],
port2eth[port_id].addr_bytes[3],
port2eth[port_id].addr_bytes[4],
port2eth[port_id].addr_bytes[5]);
*/
memset(&port_statistics, 0, sizeof(port_statistics));
}
check_all_ports_link_status(n_ports);
/* launch per-lcore init on every lcore */
rte_eal_mp_remote_launch(l2sw_launch_one_lcore, envs, CALL_MASTER);
{
uint8_t lcore_id;
RTE_LCORE_FOREACH_SLAVE(lcore_id) {
if (rte_eal_wait_lcore(lcore_id) < 0)
return -1;
}
}
rte_free(arp_table);
rte_free(route_table);
return 0;
}
int
test_distributor(void)
{
static struct rte_distributor *d;
static struct rte_mempool *p;
if (rte_lcore_count() < 2) {
printf("ERROR: not enough cores to test distributor\n");
return -1;
}
if (d == NULL) {
d = rte_distributor_create("Test_distributor", rte_socket_id(),
rte_lcore_count() - 1);
if (d == NULL) {
printf("Error creating distributor\n");
return -1;
}
} else {
rte_distributor_flush(d);
rte_distributor_clear_returns(d);
}
const unsigned nb_bufs = (511 * rte_lcore_count()) < BIG_BATCH ?
(BIG_BATCH * 2) - 1 : (511 * rte_lcore_count());
if (p == NULL) {
p = rte_mempool_create("DT_MBUF_POOL", nb_bufs,
MBUF_SIZE, BURST,
sizeof(struct rte_pktmbuf_pool_private),
rte_pktmbuf_pool_init, NULL,
rte_pktmbuf_init, NULL,
rte_socket_id(), 0);
if (p == NULL) {
printf("Error creating mempool\n");
return -1;
}
}
rte_eal_mp_remote_launch(handle_work, d, SKIP_MASTER);
if (sanity_test(d, p) < 0)
goto err;
quit_workers(d, p);
rte_eal_mp_remote_launch(handle_work_with_free_mbufs, d, SKIP_MASTER);
if (sanity_test_with_mbuf_alloc(d, p) < 0)
goto err;
quit_workers(d, p);
if (rte_lcore_count() > 2) {
rte_eal_mp_remote_launch(handle_work_for_shutdown_test, d,
SKIP_MASTER);
if (sanity_test_with_worker_shutdown(d, p) < 0)
goto err;
quit_workers(d, p);
rte_eal_mp_remote_launch(handle_work_for_shutdown_test, d,
SKIP_MASTER);
if (test_flush_with_worker_shutdown(d, p) < 0)
goto err;
quit_workers(d, p);
} else {
printf("Not enough cores to run tests for worker shutdown\n");
}
if (test_error_distributor_create_numworkers() == -1 ||
test_error_distributor_create_name() == -1) {
printf("rte_distributor_create parameter check tests failed");
return -1;
}
return 0;
err:
quit_workers(d, p);
return -1;
}
static struct vr_usocket *
usock_alloc(unsigned short proto, unsigned short type)
{
int sock_fd = -1, domain, ret;
/* socket TX buffer size = (hold flow table entries * size of jumbo frame) */
int setsocksndbuff = vr_flow_hold_limit * VR_DPDK_MAX_PACKET_SZ;
int getsocksndbuff;
socklen_t getsocksndbufflen = sizeof(getsocksndbuff);
int error = 0, flags;
unsigned int buf_len;
struct vr_usocket *usockp = NULL, *child;
bool is_socket = true;
unsigned short sock_type;
RTE_SET_USED(child);
RTE_LOG(DEBUG, USOCK, "%s[%lx]: proto %u type %u\n", __func__,
pthread_self(), proto, type);
switch (type) {
case TCP:
domain = AF_INET;
sock_type = SOCK_STREAM;
break;
case UNIX:
case RAW:
domain = AF_UNIX;
sock_type = SOCK_DGRAM;
break;
default:
return NULL;
}
if (proto == EVENT) {
is_socket = false;
sock_fd = eventfd(0, 0);
RTE_LOG(DEBUG, USOCK, "%s[%lx]: new event FD %d\n", __func__,
pthread_self(), sock_fd);
if (sock_fd < 0)
return NULL;
}
if (is_socket) {
sock_fd = socket(domain, sock_type, 0);
RTE_LOG(INFO, USOCK, "%s[%lx]: new socket FD %d\n", __func__,
pthread_self(), sock_fd);
if (sock_fd < 0)
return NULL;
/* set socket send buffer size */
ret = setsockopt(sock_fd, SOL_SOCKET, SO_SNDBUF, &setsocksndbuff,
sizeof(setsocksndbuff));
if (ret == 0) {
/* check if setting buffer succeeded */
ret = getsockopt(sock_fd, SOL_SOCKET, SO_SNDBUF, &getsocksndbuff,
&getsocksndbufflen);
if (ret == 0) {
if (getsocksndbuff >= setsocksndbuff) {
RTE_LOG(INFO, USOCK, "%s[%lx]: setting socket FD %d send buff size.\n"
"Buffer size set to %d (requested %d)\n", __func__,
pthread_self(), sock_fd, getsocksndbuff, setsocksndbuff);
} else { /* set other than requested */
RTE_LOG(ERR, USOCK, "%s[%lx]: setting socket FD %d send buff size failed.\n"
"Buffer size set to %d (requested %d)\n", __func__,
pthread_self(), sock_fd, getsocksndbuff, setsocksndbuff);
}
} else { /* requesting buffer size failed */
RTE_LOG(ERR, USOCK, "%s[%lx]: getting socket FD %d send buff size failed (%d)\n",
__func__, pthread_self(), sock_fd, errno);
}
} else { /* setting buffer size failed */
RTE_LOG(ERR, USOCK, "%s[%lx]: setting socket FD %d send buff size %d failed (%d)\n",
__func__, pthread_self(), sock_fd, setsocksndbuff, errno);
}
}
usockp = vr_zalloc(sizeof(*usockp), VR_USOCK_OBJECT);
if (!usockp)
goto error_exit;
usockp->usock_type = type;
usockp->usock_proto = proto;
usockp->usock_fd = sock_fd;
usockp->usock_state = INITED;
if (is_socket) {
error = vr_usocket_bind(usockp);
if (error < 0)
goto error_exit;
if (usockp->usock_proto == PACKET) {
error = vr_usocket_connect(usockp);
if (error < 0)
goto error_exit;
}
}
switch (proto) {
case NETLINK:
usockp->usock_max_cfds = USOCK_MAX_CHILD_FDS;
buf_len = 0;
break;
case PACKET:
usockp->usock_max_cfds = USOCK_MAX_CHILD_FDS;
buf_len = 0;
break;
case EVENT:
/* TODO: we don't need the buf since we use stack to send an event */
buf_len = USOCK_EVENT_BUF_LEN;
break;
default:
buf_len = 0;
break;
}
if (buf_len) {
usockp->usock_rx_buf = vr_zalloc(buf_len, VR_USOCK_BUF_OBJECT);
if (!usockp->usock_rx_buf)
goto error_exit;
usockp->usock_buf_len = buf_len;
usock_read_init(usockp);
}
if (proto == PACKET) {
usockp->usock_mbuf_pool = rte_mempool_lookup("packet_mbuf_pool");
if (!usockp->usock_mbuf_pool) {
usockp->usock_mbuf_pool = rte_mempool_create("packet_mbuf_pool",
PKT0_MBUF_POOL_SIZE, PKT0_MBUF_PACKET_SIZE,
PKT0_MBUF_POOL_CACHE_SZ, sizeof(struct rte_pktmbuf_pool_private),
vr_dpdk_pktmbuf_pool_init, NULL, vr_dpdk_pktmbuf_init, NULL,
rte_socket_id(), 0);
if (!usockp->usock_mbuf_pool)
goto error_exit;
}
usockp->usock_iovec = vr_zalloc(sizeof(struct iovec) *
PKT0_MAX_IOV_LEN, VR_USOCK_IOVEC_OBJECT);
if (!usockp->usock_iovec)
goto error_exit;
usock_read_init(usockp);
}
RTE_LOG(DEBUG, USOCK, "%s[%lx]: FD %d F_GETFL\n", __func__, pthread_self(),
usockp->usock_fd);
flags = fcntl(usockp->usock_fd, F_GETFL);
if (flags == -1)
goto error_exit;
RTE_LOG(DEBUG, USOCK, "%s[%lx]: FD %d F_SETFL\n", __func__, pthread_self(),
usockp->usock_fd);
error = fcntl(usockp->usock_fd, F_SETFL, flags | O_NONBLOCK);
if (error == -1)
goto error_exit;
usockp->usock_poll_block = 1;
return usockp;
error_exit:
error = errno;
if (sock_fd >= 0) {
close(sock_fd);
sock_fd = -1;
}
usock_close(usockp);
usockp = NULL;
errno = error;
return usockp;
}
fprintf(stderr, "packet get payload failed!\n");
continue;
}
ret = dpi_engine_exec(payload, len);
#endif
}
}
static int fwd_loop(__attribute__((__unused__))void *arg)
{
int lcoreid, portid, queueid;
int nb_rx, nb_tx;
int pl_len, i;
lcoreid = rte_lcore_id();
pl_len = lcore_args[lcoreid].pl_len;
printf("lcore %d from socket %d\n", rte_lcore_id(), rte_socket_id());
for(i = 0; i < pl_len; i++)
{
portid = lcore_args[lcoreid].pl[i].portid;
queueid = lcore_args[lcoreid].pl[i].queueid;
printf("lcore %d enter loop port %d queue %d\n", lcoreid, portid, queueid);
}
#ifdef EXEC_MBUF_PA_CNT
lcore_args[lcoreid].pa_ht = pacnt_hash_create(NB_MBUF << 1);
if(lcore_args[lcoreid].pa_ht == NULL)
{
rte_exit(EINVAL, "pacnt hash created failed on lcore %d\n", lcoreid);
}
#endif
while(!is_stop)
{
/* must be called per process */
int ipaugenblick_app_init(int argc,char **argv,char *app_unique_id)
{
int i;
char ringname[1024];
openlog(NULL, 0, LOG_USER);
if(rte_eal_init(argc, argv) < 0) {
syslog(LOG_ERR,"cannot initialize rte_eal");
return -1;
}
syslog(LOG_INFO,"EAL initialized\n");
free_clients_ring = rte_ring_lookup(FREE_CLIENTS_RING);
if(!free_clients_ring) {
syslog(LOG_ERR,"cannot find ring %s %d\n",__FILE__,__LINE__);
exit(0);
}
free_connections_ring = rte_ring_lookup(FREE_CONNECTIONS_RING);
if(!free_connections_ring) {
syslog(LOG_ERR,"cannot find free connections ring\n");
return -1;
}
free_connections_pool = rte_mempool_lookup(FREE_CONNECTIONS_POOL_NAME);
if(!free_connections_pool) {
syslog(LOG_ERR,"cannot find free connections pool\n");
return -1;
}
memset(local_socket_descriptors,0,sizeof(local_socket_descriptors));
for(i = 0;i < IPAUGENBLICK_CONNECTION_POOL_SIZE;i++) {
sprintf(ringname,RX_RING_NAME_BASE"%d",i);
local_socket_descriptors[i].rx_ring = rte_ring_lookup(ringname);
if(!local_socket_descriptors[i].rx_ring) {
syslog(LOG_ERR,"%s %d\n",__FILE__,__LINE__);
exit(0);
}
sprintf(ringname,TX_RING_NAME_BASE"%d",i);
local_socket_descriptors[i].tx_ring = rte_ring_lookup(ringname);
if(!local_socket_descriptors[i].tx_ring) {
syslog(LOG_ERR,"%s %d\n",__FILE__,__LINE__);
exit(0);
}
local_socket_descriptors[i].select = -1;
local_socket_descriptors[i].socket = NULL;
sprintf(ringname,"lrxcache%s_%d",app_unique_id,i);
syslog(LOG_DEBUG,"local cache name %s\n",ringname);
local_socket_descriptors[i].local_cache = rte_ring_create(ringname, 16384,rte_socket_id(), RING_F_SC_DEQ|RING_F_SP_ENQ);
if(!local_socket_descriptors[i].local_cache) {
syslog(LOG_WARNING,"cannot create local cache\n");
local_socket_descriptors[i].local_cache = rte_ring_lookup(ringname);
if(!local_socket_descriptors[i].local_cache) {
syslog(LOG_ERR,"and cannot find\n");
exit(0);
}
}
local_socket_descriptors[i].any_event_received = 0;
}
tx_bufs_pool = rte_mempool_lookup("mbufs_mempool");
if(!tx_bufs_pool) {
syslog(LOG_ERR,"cannot find tx bufs pool\n");
return -1;
}
free_command_pool = rte_mempool_lookup(FREE_COMMAND_POOL_NAME);
if(!free_command_pool) {
syslog(LOG_ERR,"cannot find free command pool\n");
return -1;
}
command_ring = rte_ring_lookup(COMMAND_RING_NAME);
if(!command_ring) {
syslog(LOG_ERR,"cannot find command ring\n");
return -1;
}
rx_bufs_ring = rte_ring_lookup("rx_mbufs_ring");
if(!rx_bufs_ring) {
syslog(LOG_ERR,"cannot find rx bufs ring\n");
return -1;
}
selectors_ring = rte_ring_lookup(SELECTOR_RING_NAME);
for(i = 0;i < IPAUGENBLICK_SELECTOR_POOL_SIZE;i++) {
sprintf(ringname,"SELECTOR_RING_NAME%d",i);
selectors[i].ready_connections = rte_ring_lookup(ringname);
if(!selectors[i].ready_connections) {
syslog(LOG_ERR,"cannot find ring %s %d\n",__FILE__,__LINE__);
exit(0);
}
}
signal(SIGHUP, sig_handler);
signal(SIGINT, sig_handler);
signal(SIGILL, sig_handler);
signal(SIGABRT, sig_handler);
signal(SIGFPE, sig_handler);
signal(SIGFPE, sig_handler);
signal(SIGSEGV, sig_handler);
signal(SIGTERM, sig_handler);
signal(SIGUSR1, sig_handler);
// pthread_create(&stats_thread,NULL,print_stats,NULL);
return ((tx_bufs_pool == NULL)||(command_ring == NULL)||(free_command_pool == NULL));
}
static int configure_eth_port(int32_t id) {
struct port_configure* pconf = &ports_conf[id];
struct ether_addr* addr = &pconf->addr;
const uint16_t rxRings = 1, txRings = 1;
const uint8_t nb_ports = rte_eth_dev_count();
int ret;
uint16_t q;
int32_t port_id = pconf->portid;
char buf[1024];
pconf->out_buf.count = 0;
if (port_id > nb_ports)
return -1;
ret = rte_eth_dev_configure(port_id, rxRings, txRings, &port_conf_default);
if (ret != 0)
return ret;
for (q = 0; q < rxRings; q++) {
ret = rte_eth_rx_queue_setup(port_id, q, RX_DESC_PER_QUEUE,
rte_eth_dev_socket_id(port_id), NULL, mbuf_pool);
if (ret < 0)
return ret;
}
for (q = 0; q < txRings; q++) {
ret = rte_eth_tx_queue_setup(port_id, q, TX_DESC_PER_QUEUE,
rte_eth_dev_socket_id(port_id), NULL);
if (ret < 0)
return ret;
}
snprintf(buf, sizeof(buf), "ring_in_%d", port_id);
pconf->ring_in = rte_ring_create(buf, RING_SIZE, rte_socket_id(),
RING_F_SP_ENQ);
if (pconf->ring_in == NULL) {
return -1;
}
snprintf(buf, sizeof(buf), "ring_out_%d", port_id);
pconf->ring_out = rte_ring_create(buf, RING_SIZE, rte_socket_id(),
RING_F_SC_DEQ);
if (pconf->ring_out == NULL) {
return -1;
}
ret = rte_eth_dev_start(port_id);
if (ret < 0)
return ret;
rte_eth_macaddr_get(port_id, addr);
printf("Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
" %02"PRIx8" %02"PRIx8" %02"PRIx8"\n", (unsigned) port_id,
addr->addr_bytes[0], addr->addr_bytes[1], addr->addr_bytes[2],
addr->addr_bytes[3], addr->addr_bytes[4], addr->addr_bytes[5]);
rte_eth_promiscuous_enable(port_id);
return 0;
}
