#define CMD_LINE_OPT_NB_QUEUE "nb-queue"
#define CMD_LINE_OPT_COPY_TYPE "copy-type"
#define CMD_LINE_OPT_RING_SIZE "ring-size"
+#define CMD_LINE_OPT_BATCH_SIZE "dma-batch-size"
+#define CMD_LINE_OPT_FRAME_SIZE "max-frame-size"
+#define CMD_LINE_OPT_STATS_INTERVAL "stats-interval"
/* configurable number of RX/TX ring descriptors */
#define RX_DEFAULT_RINGSIZE 1024
*/
static unsigned short ring_size = 2048;
+/* interval, in seconds, between stats prints */
+static unsigned short stats_interval = 1;
+
/* global transmission config */
struct rxtx_transmission_config cfg;
static volatile bool force_quit;
+static uint32_t ioat_batch_sz = MAX_PKT_BURST;
+static uint32_t max_frame_size = RTE_ETHER_MAX_LEN;
+
/* ethernet addresses of ports */
static struct rte_ether_addr ioat_ports_eth_addr[RTE_MAX_ETHPORTS];
"\nTotal packets Tx: %24"PRIu64" [pps]"
"\nTotal packets Rx: %24"PRIu64" [pps]"
"\nTotal packets dropped: %19"PRIu64" [pps]",
- ts->total_packets_tx,
- ts->total_packets_rx,
- ts->total_packets_dropped);
+ ts->total_packets_tx / stats_interval,
+ ts->total_packets_rx / stats_interval,
+ ts->total_packets_dropped / stats_interval);
if (copy_mode == COPY_MODE_IOAT_NUM) {
printf("\nTotal IOAT successful enqueues: %8"PRIu64" [enq/s]"
"\nTotal IOAT failed enqueues: %12"PRIu64" [enq/s]",
- ts->total_successful_enqueues,
- ts->total_failed_enqueues);
+ ts->total_successful_enqueues / stats_interval,
+ ts->total_failed_enqueues / stats_interval);
}
printf("\n====================================================\n");
memset(&ts, 0, sizeof(struct total_statistics));
while (!force_quit) {
- /* Sleep for 1 second each round - init sleep allows reading
+ /* Sleep for "stats_interval" seconds each round - init sleep allows reading
* messages from app startup.
*/
- sleep(1);
+ sleep(stats_interval);
/* Clear screen and move to top left */
printf("%s%s", clr, topLeft);
/* Perform packet copy there is a user-defined function. 8< */
static inline void
-pktmbuf_sw_copy(struct rte_mbuf *src, struct rte_mbuf *dst)
+pktmbuf_metadata_copy(const struct rte_mbuf *src, struct rte_mbuf *dst)
{
- /* Copy packet metadata */
- rte_memcpy(&dst->rearm_data,
- &src->rearm_data,
- offsetof(struct rte_mbuf, cacheline1)
- - offsetof(struct rte_mbuf, rearm_data));
+ dst->data_off = src->data_off;
+ memcpy(&dst->rx_descriptor_fields1, &src->rx_descriptor_fields1,
+ offsetof(struct rte_mbuf, buf_len) -
+ offsetof(struct rte_mbuf, rx_descriptor_fields1));
+}
- /* Copy packet data */
+/* Copy packet data */
+static inline void
+pktmbuf_sw_copy(struct rte_mbuf *src, struct rte_mbuf *dst)
+{
rte_memcpy(rte_pktmbuf_mtod(dst, char *),
rte_pktmbuf_mtod(src, char *), src->data_len);
}
/* >8 End of perform packet copy there is a user-defined function. */
static uint32_t
-ioat_enqueue_packets(struct rte_mbuf **pkts,
+ioat_enqueue_packets(struct rte_mbuf *pkts[], struct rte_mbuf *pkts_copy[],
uint32_t nb_rx, uint16_t dev_id)
{
int ret;
uint32_t i;
- struct rte_mbuf *pkts_copy[MAX_PKT_BURST];
-
- const uint64_t addr_offset = RTE_PTR_DIFF(pkts[0]->buf_addr,
- &pkts[0]->rearm_data);
-
- ret = rte_mempool_get_bulk(ioat_pktmbuf_pool,
- (void *)pkts_copy, nb_rx);
-
- if (unlikely(ret < 0))
- rte_exit(EXIT_FAILURE, "Unable to allocate memory.\n");
for (i = 0; i < nb_rx; i++) {
/* Perform data copy */
ret = rte_ioat_enqueue_copy(dev_id,
- pkts[i]->buf_iova - addr_offset,
- pkts_copy[i]->buf_iova - addr_offset,
- rte_pktmbuf_data_len(pkts[i]) + addr_offset,
+ rte_pktmbuf_iova(pkts[i]),
+ rte_pktmbuf_iova(pkts_copy[i]),
+ rte_pktmbuf_data_len(pkts[i]),
(uintptr_t)pkts[i],
(uintptr_t)pkts_copy[i]);
}
ret = i;
- /* Free any not enqueued packets. */
- rte_mempool_put_bulk(ioat_pktmbuf_pool, (void *)&pkts[i], nb_rx - i);
- rte_mempool_put_bulk(ioat_pktmbuf_pool, (void *)&pkts_copy[i],
- nb_rx - i);
-
return ret;
}
+static inline uint32_t
+ioat_enqueue(struct rte_mbuf *pkts[], struct rte_mbuf *pkts_copy[],
+ uint32_t num, uint32_t step, uint16_t dev_id)
+{
+ uint32_t i, k, m, n;
+
+ k = 0;
+ for (i = 0; i < num; i += m) {
+
+ m = RTE_MIN(step, num - i);
+ n = ioat_enqueue_packets(pkts + i, pkts_copy + i, m, dev_id);
+ k += n;
+ if (n > 0)
+ rte_ioat_perform_ops(dev_id);
+
+ /* don't try to enqueue more if HW queue is full */
+ if (n != m)
+ break;
+ }
+
+ return k;
+}
+
+static inline uint32_t
+ioat_dequeue(struct rte_mbuf *src[], struct rte_mbuf *dst[], uint32_t num,
+ uint16_t dev_id)
+{
+ int32_t rc;
+ /* Dequeue the mbufs from IOAT device. Since all memory
+ * is DPDK pinned memory and therefore all addresses should
+ * be valid, we don't check for copy errors
+ */
+ rc = rte_ioat_completed_ops(dev_id, num, NULL, NULL,
+ (void *)src, (void *)dst);
+ if (rc < 0) {
+ RTE_LOG(CRIT, IOAT,
+ "rte_ioat_completed_ops(%hu) failedi, error: %d\n",
+ dev_id, rte_errno);
+ rc = 0;
+ }
+ return rc;
+}
+
/* Receive packets on one port and enqueue to IOAT rawdev or rte_ring. 8< */
static void
ioat_rx_port(struct rxtx_port_config *rx_config)
{
+ int32_t ret;
uint32_t nb_rx, nb_enq, i, j;
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+ struct rte_mbuf *pkts_burst_copy[MAX_PKT_BURST];
for (i = 0; i < rx_config->nb_queues; i++) {
port_statistics.rx[rx_config->rxtx_port] += nb_rx;
+ ret = rte_mempool_get_bulk(ioat_pktmbuf_pool,
+ (void *)pkts_burst_copy, nb_rx);
+
+ if (unlikely(ret < 0))
+ rte_exit(EXIT_FAILURE,
+ "Unable to allocate memory.\n");
+
+ for (j = 0; j < nb_rx; j++)
+ pktmbuf_metadata_copy(pkts_burst[j],
+ pkts_burst_copy[j]);
+
if (copy_mode == COPY_MODE_IOAT_NUM) {
- /* Perform packet hardware copy */
- nb_enq = ioat_enqueue_packets(pkts_burst,
- nb_rx, rx_config->ioat_ids[i]);
- if (nb_enq > 0)
- rte_ioat_perform_ops(rx_config->ioat_ids[i]);
- } else {
- /* Perform packet software copy, free source packets */
- int ret;
- struct rte_mbuf *pkts_burst_copy[MAX_PKT_BURST];
- ret = rte_mempool_get_bulk(ioat_pktmbuf_pool,
- (void *)pkts_burst_copy, nb_rx);
+ /* enqueue packets for hardware copy */
+ nb_enq = ioat_enqueue(pkts_burst, pkts_burst_copy,
+ nb_rx, ioat_batch_sz, rx_config->ioat_ids[i]);
+
+ /* free any not enqueued packets. */
+ rte_mempool_put_bulk(ioat_pktmbuf_pool,
+ (void *)&pkts_burst[nb_enq],
+ nb_rx - nb_enq);
+ rte_mempool_put_bulk(ioat_pktmbuf_pool,
+ (void *)&pkts_burst_copy[nb_enq],
+ nb_rx - nb_enq);
- if (unlikely(ret < 0))
- rte_exit(EXIT_FAILURE,
- "Unable to allocate memory.\n");
+ port_statistics.copy_dropped[rx_config->rxtx_port] +=
+ (nb_rx - nb_enq);
+ /* get completed copies */
+ nb_rx = ioat_dequeue(pkts_burst, pkts_burst_copy,
+ MAX_PKT_BURST, rx_config->ioat_ids[i]);
+ } else {
+ /* Perform packet software copy, free source packets */
for (j = 0; j < nb_rx; j++)
pktmbuf_sw_copy(pkts_burst[j],
pkts_burst_copy[j]);
+ }
- rte_mempool_put_bulk(ioat_pktmbuf_pool,
- (void *)pkts_burst, nb_rx);
+ rte_mempool_put_bulk(ioat_pktmbuf_pool,
+ (void *)pkts_burst, nb_rx);
- nb_enq = rte_ring_enqueue_burst(
- rx_config->rx_to_tx_ring,
- (void *)pkts_burst_copy, nb_rx, NULL);
+ nb_enq = rte_ring_enqueue_burst(rx_config->rx_to_tx_ring,
+ (void *)pkts_burst_copy, nb_rx, NULL);
- /* Free any not enqueued packets. */
- rte_mempool_put_bulk(ioat_pktmbuf_pool,
- (void *)&pkts_burst_copy[nb_enq],
- nb_rx - nb_enq);
- }
+ /* Free any not enqueued packets. */
+ rte_mempool_put_bulk(ioat_pktmbuf_pool,
+ (void *)&pkts_burst_copy[nb_enq],
+ nb_rx - nb_enq);
port_statistics.copy_dropped[rx_config->rxtx_port] +=
(nb_rx - nb_enq);
static void
ioat_tx_port(struct rxtx_port_config *tx_config)
{
- uint32_t i, j, nb_dq = 0;
- struct rte_mbuf *mbufs_src[MAX_PKT_BURST];
- struct rte_mbuf *mbufs_dst[MAX_PKT_BURST];
+ uint32_t i, j, nb_dq, nb_tx;
+ struct rte_mbuf *mbufs[MAX_PKT_BURST];
for (i = 0; i < tx_config->nb_queues; i++) {
- if (copy_mode == COPY_MODE_IOAT_NUM) {
- /* Dequeue the mbufs from IOAT device. Since all memory
- * is DPDK pinned memory and therefore all addresses should
- * be valid, we don't check for copy errors
- */
- nb_dq = rte_ioat_completed_ops(
- tx_config->ioat_ids[i], MAX_PKT_BURST, NULL, NULL,
- (void *)mbufs_src, (void *)mbufs_dst);
- } else {
- /* Dequeue the mbufs from rx_to_tx_ring. */
- nb_dq = rte_ring_dequeue_burst(
- tx_config->rx_to_tx_ring, (void *)mbufs_dst,
- MAX_PKT_BURST, NULL);
- }
-
- if ((int32_t) nb_dq <= 0)
- return;
- if (copy_mode == COPY_MODE_IOAT_NUM)
- rte_mempool_put_bulk(ioat_pktmbuf_pool,
- (void *)mbufs_src, nb_dq);
+ /* Dequeue the mbufs from rx_to_tx_ring. */
+ nb_dq = rte_ring_dequeue_burst(tx_config->rx_to_tx_ring,
+ (void *)mbufs, MAX_PKT_BURST, NULL);
+ if (nb_dq == 0)
+ continue;
/* Update macs if enabled */
if (mac_updating) {
for (j = 0; j < nb_dq; j++)
- update_mac_addrs(mbufs_dst[j],
+ update_mac_addrs(mbufs[j],
tx_config->rxtx_port);
}
- const uint16_t nb_tx = rte_eth_tx_burst(
- tx_config->rxtx_port, 0,
- (void *)mbufs_dst, nb_dq);
+ nb_tx = rte_eth_tx_burst(tx_config->rxtx_port, 0,
+ (void *)mbufs, nb_dq);
port_statistics.tx[tx_config->rxtx_port] += nb_tx;
/* Free any unsent packets. */
if (unlikely(nb_tx < nb_dq))
rte_mempool_put_bulk(ioat_pktmbuf_pool,
- (void *)&mbufs_dst[nb_tx],
- nb_dq - nb_tx);
+ (void *)&mbufs[nb_tx], nb_dq - nb_tx);
}
}
/* >8 End of transmitting packets from IOAT. */
ioat_usage(const char *prgname)
{
printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
+ " -b --dma-batch-size: number of requests per DMA batch\n"
+ " -f --max-frame-size: max frame size\n"
" -p --portmask: hexadecimal bitmask of ports to configure\n"
" -q NQ: number of RX queues per port (default is 1)\n"
" --[no-]mac-updating: Enable or disable MAC addresses updating (enabled by default)\n"
" - The source MAC address is replaced by the TX port MAC address\n"
" - The destination MAC address is replaced by 02:00:00:00:00:TX_PORT_ID\n"
" -c --copy-type CT: type of copy: sw|hw\n"
- " -s --ring-size RS: size of IOAT rawdev ring for hardware copy mode or rte_ring for software copy mode\n",
+ " -s --ring-size RS: size of IOAT rawdev ring for hardware copy mode or rte_ring for software copy mode\n"
+ " -i --stats-interval SI: interval, in seconds, between stats prints (default is 1)\n",
prgname);
}
ioat_parse_args(int argc, char **argv, unsigned int nb_ports)
{
static const char short_options[] =
+ "b:" /* dma batch size */
+ "c:" /* copy type (sw|hw) */
+ "f:" /* max frame size */
"p:" /* portmask */
"q:" /* number of RX queues per port */
- "c:" /* copy type (sw|hw) */
"s:" /* ring size */
+ "i:" /* interval, in seconds, between stats prints */
;
static const struct option lgopts[] = {
{CMD_LINE_OPT_NB_QUEUE, required_argument, NULL, 'q'},
{CMD_LINE_OPT_COPY_TYPE, required_argument, NULL, 'c'},
{CMD_LINE_OPT_RING_SIZE, required_argument, NULL, 's'},
+ {CMD_LINE_OPT_BATCH_SIZE, required_argument, NULL, 'b'},
+ {CMD_LINE_OPT_FRAME_SIZE, required_argument, NULL, 'f'},
+ {CMD_LINE_OPT_STATS_INTERVAL, required_argument, NULL, 'i'},
{NULL, 0, 0, 0}
};
lgopts, &option_index)) != EOF) {
switch (opt) {
+ case 'b':
+ ioat_batch_sz = atoi(optarg);
+ if (ioat_batch_sz > MAX_PKT_BURST) {
+ printf("Invalid dma batch size, %s.\n", optarg);
+ ioat_usage(prgname);
+ return -1;
+ }
+ break;
+ case 'f':
+ max_frame_size = atoi(optarg);
+ if (max_frame_size > RTE_ETHER_MAX_JUMBO_FRAME_LEN) {
+ printf("Invalid max frame size, %s.\n", optarg);
+ ioat_usage(prgname);
+ return -1;
+ }
+ break;
+
/* portmask */
case 'p':
ioat_enabled_port_mask = ioat_parse_portmask(optarg);
}
break;
+ case 'i':
+ stats_interval = atoi(optarg);
+ if (stats_interval == 0) {
+ printf("Invalid stats interval, setting to 1\n");
+ stats_interval = 1; /* set to default */
+ }
+ break;
+
/* long options */
case 0:
break;
/* Configuring port to use RSS for multiple RX queues. 8< */
static const struct rte_eth_conf port_conf = {
.rxmode = {
- .mq_mode = ETH_MQ_RX_RSS,
+ .mq_mode = RTE_ETH_MQ_RX_RSS,
},
.rx_adv_conf = {
.rss_conf = {
.rss_key = NULL,
- .rss_hf = ETH_RSS_PROTO_MASK,
+ .rss_hf = RTE_ETH_RSS_PROTO_MASK,
}
}
};
struct rte_eth_dev_info dev_info;
int ret, i;
+ if (max_frame_size > local_port_conf.rxmode.mtu)
+ local_port_conf.rxmode.mtu = max_frame_size;
+
/* Skip ports that are not enabled */
if ((ioat_enabled_port_mask & (1 << portid)) == 0) {
printf("Skipping disabled port %u\n", portid);
local_port_conf.rx_adv_conf.rss_conf.rss_hf &=
dev_info.flow_type_rss_offloads;
- if (dev_info.tx_offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE)
- local_port_conf.txmode.offloads |=
- DEV_TX_OFFLOAD_MBUF_FAST_FREE;
ret = rte_eth_dev_configure(portid, nb_queues, 1, &local_port_conf);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Cannot configure device:"
cfg.ports[cfg.nb_ports++].nb_queues = nb_queues;
}
+/* Get a device dump for each device being used by the application */
+static void
+rawdev_dump(void)
+{
+ uint32_t i, j;
+
+ if (copy_mode != COPY_MODE_IOAT_NUM)
+ return;
+
+ for (i = 0; i < cfg.nb_ports; i++)
+ for (j = 0; j < cfg.ports[i].nb_queues; j++)
+ rte_rawdev_dump(cfg.ports[i].ioat_ids[j], stdout);
+}
+
static void
signal_handler(int signum)
{
printf("\n\nSignal %d received, preparing to exit...\n",
signum);
force_quit = true;
+ } else if (signum == SIGUSR1) {
+ rawdev_dump();
}
}
uint16_t nb_ports, portid;
uint32_t i;
unsigned int nb_mbufs;
+ size_t sz;
/* Init EAL. 8< */
ret = rte_eal_init(argc, argv);
force_quit = false;
signal(SIGINT, signal_handler);
signal(SIGTERM, signal_handler);
+ signal(SIGUSR1, signal_handler);
nb_ports = rte_eth_dev_count_avail();
if (nb_ports == 0)
/* Allocates mempool to hold the mbufs. 8< */
nb_mbufs = RTE_MAX(nb_ports * (nb_queues * (nb_rxd + nb_txd +
- 4 * MAX_PKT_BURST) + rte_lcore_count() * MEMPOOL_CACHE_SIZE),
+ 4 * MAX_PKT_BURST + ring_size) + ring_size +
+ rte_lcore_count() * MEMPOOL_CACHE_SIZE),
MIN_POOL_SIZE);
/* Create the mbuf pool */
+ sz = max_frame_size + RTE_PKTMBUF_HEADROOM;
+ sz = RTE_MAX(sz, (size_t)RTE_MBUF_DEFAULT_BUF_SIZE);
ioat_pktmbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", nb_mbufs,
- MEMPOOL_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
- rte_socket_id());
+ MEMPOOL_CACHE_SIZE, 0, sz, rte_socket_id());
if (ioat_pktmbuf_pool == NULL)
rte_exit(EXIT_FAILURE, "Cannot init mbuf pool\n");
/* >8 End of allocates mempool to hold the mbufs. */
if (copy_mode == COPY_MODE_IOAT_NUM)
assign_rawdevs();
- else /* copy_mode == COPY_MODE_SW_NUM */
- assign_rings();
+
+ assign_rings();
/* >8 End of assigning each port resources. */
start_forwarding_cores();
git.droids-corp.org / dpdk.git / blobdiff