Re: [dpdk-dev] [PATCH v2] af_packet: Fix some klocwork errors

[Neil Horman <nhorman@tuxdriver.com>]

On Fri, Feb 27, 2015 at 08:49:13AM +0800, Ouyang Changchun wrote:
> Fix possible memory leak issue: free kvlist before return;
> Fix possible resource lost issue: close qssockfd before return;
> 
> Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
> ---
> change in v2:
>   - Make the error exit point a common path.
> 
>  lib/librte_pmd_af_packet/rte_eth_af_packet.c | 11 +++++++++--
>  1 file changed, 9 insertions(+), 2 deletions(-)
> 
> diff --git a/lib/librte_pmd_af_packet/rte_eth_af_packet.c b/lib/librte_pmd_af_packet/rte_eth_af_packet.c
> index 80e9bdf..c675724 100644
> --- a/lib/librte_pmd_af_packet/rte_eth_af_packet.c
> +++ b/lib/librte_pmd_af_packet/rte_eth_af_packet.c
> @@ -694,6 +694,8 @@ error:
>  		}
>  		rte_free(*internals);
>  	}
> +	if (qsockfd != -1)
> +		close(qsockfd);
This is insufficient.  qsockfd is a loop index that is reassigned for each tx
queue we have.  In the error path you need to do this, and loop through the tx
queues, closing each tx_queue->sockfd.

>  	return -1;
>  }
>  
> @@ -823,16 +825,21 @@ rte_pmd_af_packet_devinit(const char *name, const char *params)
>  		ret = rte_kvargs_process(kvlist, ETH_AF_PACKET_IFACE_ARG,
>  		                         &open_packet_iface, &sockfd);
>  		if (ret < 0)
> -			return -1;
> +			goto error;
>  	}
>  
>  	ret = rte_eth_from_packet(name, &sockfd, numa_node, kvlist);
>  	close(sockfd); /* no longer needed */
>  
>  	if (ret < 0)
> -		return -1;
> +		goto error;
>  
> +	rte_kvargs_free(kvlist);
>  	return 0;
> +
> +error:
> +	rte_kvargs_free(kvlist);
> +	return -1;
>  }
>  
>  static struct rte_driver pmd_af_packet_drv = {
> -- 
> 1.8.4.2
> 
>