Re: [PATCH] netpoll: Add support for hardware checksumming on egress

[Eric Dumazet <eric.dumazet@gmail.com>]

On Tue, 2012-06-12 at 13:26 +0300, Bogdan Hamciuc wrote:
> Netpoll used to compute its own csum; but if the device supports, we
> should let it do the checksum itself.
> 
> Signed-off-by: Bogdan Hamciuc <bogdan.hamciuc@freescale.com>
> ---
>  net/core/netpoll.c |   14 ++++++++++----
>  1 files changed, 10 insertions(+), 4 deletions(-)
> 
> diff --git a/net/core/netpoll.c b/net/core/netpoll.c
> index 9a08068..f5d00b4 100644
> --- a/net/core/netpoll.c
> +++ b/net/core/netpoll.c
> @@ -385,13 +385,19 @@ void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
>  	udph->source = htons(np->local_port);
>  	udph->dest = htons(np->remote_port);
>  	udph->len = htons(udp_len);
> -	udph->check = 0;
> -	udph->check = csum_tcpudp_magic(np->local_ip,
> +
> +	/* Only querying the IPv4 csumming capabilities */
> +	if (np->dev->features & NETIF_F_IP_CSUM)
> +		skb->ip_summed = CHECKSUM_PARTIAL;
> +	else {
> +		skb->ip_summed = CHECKSUM_NONE;
> +		udph->check = csum_tcpudp_magic(np->local_ip,
>  					np->remote_ip,
>  					udp_len, IPPROTO_UDP,
>  					csum_partial(udph, udp_len, 0));
> -	if (udph->check == 0)
> -		udph->check = CSUM_MANGLED_0;
> +		if (udph->check == 0)
> +			udph->check = CSUM_MANGLED_0;
> +	}
>  
>  	skb_push(skb, sizeof(*iph));
>  	skb_reset_network_header(skb);

Hi Bogdan

I cant see how this can possibly work ?

Which NIC was able to send a good UDP frame after this patch ?