Re: [PATCH v2] virtio: Fix crash issue for secondary process

[Thomas Monjalon <thomas.monjalon-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>]

2015-03-27 14:35, Ouyang Changchun:
> It needs Rx function even in the case of secondary process, and it also needs check if
> it supports mergeable feature or not.
> 
> Signed-off-by: Changchun Ouyang <changchun.ouyang-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> ---
> 
> Changes in v2:
>   -- Check if it supports mergeable or not for the secondary process.

> --- a/lib/librte_pmd_virtio/virtio_ethdev.c
> +++ b/lib/librte_pmd_virtio/virtio_ethdev.c
> @@ -1115,8 +1115,13 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
>  	eth_dev->dev_ops = &virtio_eth_dev_ops;
>  	eth_dev->tx_pkt_burst = &virtio_xmit_pkts;
>  
> -	if (rte_eal_process_type() == RTE_PROC_SECONDARY)
> +	if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
> +		if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF))
> +			eth_dev->rx_pkt_burst = &virtio_recv_mergeable_pkts;
> +		else
> +			eth_dev->rx_pkt_burst = &virtio_recv_pkts;

You are duplicating code, making it error prone for later maintenance.
Please merge primary and secondary cases above the if() block.

>  		return 0;
> +	}
>  
>  	/* Allocate memory for storing MAC addresses */
>  	eth_dev->data->mac_addrs = rte_zmalloc("virtio", ETHER_ADDR_LEN, 0);
>