From: Chengwen Feng <fengcheng...@huawei.com>
In the reset restore-conf phase, the reset process will allocate for
the Rx ring mbufs unconditionlly.
And the rte_eth_dev_rx_queue_start() will also allocate for the Rx ring
mbufs unconditionlly.
So if the rte_eth_dev_rx_queue_start() is invoked after restore-conf
phase, then the mbufs allocated in restore-conf phase will leak.
So fix it by conditional release Rx ring mbufs in
rte_eth_dev_rx_queue_start(): if the Rx ring mbufs were allocated then
release them first.
This patch also set all sw-ring[]'s mbuf is NULL when release Rx ring
mbufs so that we can determine whether the Rx ring mbufs were allocated
based only on the first sw-ring[0]'s mbuf.
Fixes: fa29fe45a7b4 ("net/hns3: support queue start and stop")
Cc: sta...@dpdk.org
Signed-off-by: Chengwen Feng <fengcheng...@huawei.com>
Signed-off-by: Dongdong Liu <liudongdo...@huawei.com>
---
drivers/net/hns3/hns3_rxtx.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/drivers/net/hns3/hns3_rxtx.c b/drivers/net/hns3/hns3_rxtx.c
index 2bfc5507e3..2493748683 100644
--- a/drivers/net/hns3/hns3_rxtx.c
+++ b/drivers/net/hns3/hns3_rxtx.c
@@ -50,6 +50,8 @@ hns3_rx_queue_release_mbufs(struct hns3_rx_queue *rxq)
rxq->sw_ring[i].mbuf = NULL;
}
}
+ for (i = 0; i < rxq->rx_rearm_nb; i++)
+ rxq->sw_ring[rxq->rx_rearm_start + i].mbuf = NULL;
}
for (i = 0; i < rxq->bulk_mbuf_num; i++)
@@ -4538,6 +4540,9 @@ hns3_dev_rx_queue_start(struct rte_eth_dev *dev, uint16_t
rx_queue_id)
return ret;
}
+ if (rxq->sw_ring[0].mbuf != NULL)
+ hns3_rx_queue_release_mbufs(rxq);
+
ret = hns3_init_rxq(hns, rx_queue_id);
if (ret) {
hns3_err(hw, "fail to init Rx queue %u, ret = %d.",
--
2.22.0
