The statistics logging are in the data path, as rte_atomic APIs enforce
unnecessary barriers on aarch64, this patch uses c11 atomic APIs with
the __ATOMIC_RELAXED memory ordering to give CPU more freedom to optimize
by making use of out-of-order execution.
Signed-off-by: Gavin Hu <gavin.hu@arm.com>
Reviewed-by: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
---
drivers/net/ena/ena_ethdev.c | 38 ++++++++++++++++++++++++------------
drivers/net/ena/ena_ethdev.h | 8 ++++----
2 files changed, 29 insertions(+), 17 deletions(-)
@@ -8,7 +8,6 @@
#include <rte_ethdev_driver.h>
#include <rte_ethdev_pci.h>
#include <rte_tcp.h>
-#include <rte_atomic.h>
#include <rte_dev.h>
#include <rte_errno.h>
#include <rte_version.h>
@@ -902,10 +901,14 @@ static void ena_stats_restart(struct rte_eth_dev *dev)
{
struct ena_adapter *adapter = dev->data->dev_private;
- rte_atomic64_init(&adapter->drv_stats->ierrors);
- rte_atomic64_init(&adapter->drv_stats->oerrors);
- rte_atomic64_init(&adapter->drv_stats->rx_nombuf);
- rte_atomic64_init(&adapter->drv_stats->rx_drops);
+ __atomic_store_n(&adapter->drv_stats->ierrors, 1,
+ __ATOMIC_RELAXED);
+ __atomic_store_n(&adapter->drv_stats->oerrors, 1,
+ __ATOMIC_RELAXED);
+ __atomic_store_n(&adapter->drv_stats->rx_nombuf, 1,
+ __ATOMIC_RELAXED);
+ __atomic_store_n(&adapter->drv_stats->rx_drops, 1,
+ __ATOMIC_RELAXED);
}
static int ena_stats_get(struct rte_eth_dev *dev,
@@ -939,10 +942,14 @@ static int ena_stats_get(struct rte_eth_dev *dev,
ena_stats.tx_bytes_low);
/* Driver related stats */
- stats->imissed = rte_atomic64_read(&adapter->drv_stats->rx_drops);
- stats->ierrors = rte_atomic64_read(&adapter->drv_stats->ierrors);
- stats->oerrors = rte_atomic64_read(&adapter->drv_stats->oerrors);
- stats->rx_nombuf = rte_atomic64_read(&adapter->drv_stats->rx_nombuf);
+ stats->imissed = __atomic_load_n(&adapter->drv_stats->rx_drops,
+ __ATOMIC_RELAXED);
+ stats->ierrors = __atomic_load_n(&adapter->drv_stats->ierrors,
+ __ATOMIC_RELAXED);
+ stats->oerrors = __atomic_load_n(&adapter->drv_stats->oerrors,
+ __ATOMIC_RELAXED);
+ stats->rx_nombuf = __atomic_load_n(&adapter->drv_stats->rx_nombuf,
+ __ATOMIC_RELAXED);
max_rings_stats = RTE_MIN(dev->data->nb_rx_queues,
RTE_ETHDEV_QUEUE_STAT_CNTRS);
@@ -1376,7 +1383,8 @@ static int ena_populate_rx_queue(struct ena_ring *rxq, unsigned int count)
/* get resources for incoming packets */
rc = rte_mempool_get_bulk(rxq->mb_pool, (void **)mbufs, count);
if (unlikely(rc < 0)) {
- rte_atomic64_inc(&rxq->adapter->drv_stats->rx_nombuf);
+ __atomic_add_fetch(&rxq->adapter->drv_stats->rx_nombuf, 1,
+ __ATOMIC_RELAXED);
++rxq->rx_stats.mbuf_alloc_fail;
PMD_RX_LOG(DEBUG, "there are no enough free buffers");
return 0;
@@ -2074,7 +2082,9 @@ static uint16_t eth_ena_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
if (unlikely(mbuf_head->ol_flags &
(PKT_RX_IP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD))) {
- rte_atomic64_inc(&rx_ring->adapter->drv_stats->ierrors);
+ __atomic_add_fetch
+ (&rx_ring->adapter->drv_stats->ierrors,
+ 1, __ATOMIC_RELAXED);
++rx_ring->rx_stats.bad_csum;
}
@@ -2214,7 +2224,8 @@ static int ena_check_and_linearize_mbuf(struct ena_ring *tx_ring,
rc = rte_pktmbuf_linearize(mbuf);
if (unlikely(rc)) {
PMD_DRV_LOG(WARNING, "Mbuf linearize failed\n");
- rte_atomic64_inc(&tx_ring->adapter->drv_stats->ierrors);
+ __atomic_add_fetch(&tx_ring->adapter->drv_stats->ierrors,
+ 1, __ATOMIC_RELAXED);
++tx_ring->tx_stats.linearize_failed;
return rc;
}
@@ -2682,7 +2693,8 @@ static void ena_keep_alive(void *adapter_data,
desc = (struct ena_admin_aenq_keep_alive_desc *)aenq_e;
rx_drops = ((uint64_t)desc->rx_drops_high << 32) | desc->rx_drops_low;
- rte_atomic64_set(&adapter->drv_stats->rx_drops, rx_drops);
+ __atomic_store_n(&adapter->drv_stats->rx_drops, rx_drops,
+ __ATOMIC_RELAXED);
}
/**
@@ -129,10 +129,10 @@ enum ena_adapter_state {
};
struct ena_driver_stats {
- rte_atomic64_t ierrors;
- rte_atomic64_t oerrors;
- rte_atomic64_t rx_nombuf;
- rte_atomic64_t rx_drops;
+ uint64_t ierrors;
+ uint64_t oerrors;
+ uint64_t rx_nombuf;
+ uint64_t rx_drops;
};
struct ena_stats_dev {