napatech / dpdk Goto Github PK

When rte_flow_create fails due to an error in DoNtpl, the next DPDK call can hang due to a deadlock.
In my case I tried to clean up after the failed rte_flow_create by calling rte_flow_flush, which hangs indefinitely (waiting for a lock).

I believe the cause is a missing NTACC_UNLOCK(&internals->configlock); when an error occurs in DoNtpl:

We can store the pcap with pdump, if we are in a need of dissection, is there any good way to do it?

Hi, developers, thank you for your checking. The lock &sh->txpp.mutex may be not released correctly the method returns at the branch !sh->txpp.refcnt || --sh->txpp.refcnt satisfies. The fix is to insert pthread_mutex_lock(&sh->txpp.mutex); before returning.

void
mlx5_txpp_stop(struct rte_eth_dev *dev)
{
	struct mlx5_priv *priv = dev->data->dev_private;
	struct mlx5_dev_ctx_shared *sh = priv->sh;
	int ret;

	if (!priv->txpp_en) {
		return;
	}
	priv->txpp_en = 0;
	ret = pthread_mutex_lock(&sh->txpp.mutex); // the lock
	MLX5_ASSERT(!ret);
	RTE_SET_USED(ret);
	MLX5_ASSERT(sh->txpp.refcnt);
	if (!sh->txpp.refcnt || --sh->txpp.refcnt)
		return;  // return without releasing
	
	mlx5_txpp_destroy(sh);
	ret = pthread_mutex_unlock(&sh->txpp.mutex);
	MLX5_ASSERT(!ret);
	RTE_SET_USED(ret);
}

Hi, developers, I found a lock acquisition with releasing in the function createoptimizedFilter. Is it ture? Thank you for your checking.