> + /* code added for copybreak, this should improve
> + * performance for small packets with large amounts
> + * of reassembly being done in the stack */
> +#define IXGB_CB_LENGTH 256
> + if (length < IXGB_CB_LENGTH) {
> + struct sk_buff *new_skb =
> + dev_alloc_skb(length + NET_IP_ALIGN);
> + if (new_skb) {
> + skb_reserve(new_skb, NET_IP_ALIGN);
> + new_skb->dev = netdev;
> + memcpy(new_skb->data - NET_IP_ALIGN,
> + skb->data - NET_IP_ALIGN,
> + length + NET_IP_ALIGN);
> + /* save the skb in buffer_info as good */
> + buffer_info->skb = skb;
> + skb = new_skb;
> + }
> + }
>
Why use dev_alloc_skb in the receive path? All it does is add
an extra 16 bytes of header space that is never used.
Also, you should use 2 rather than NET_IP_ALIGN here.
The point of NET_IP_ALIGN is to allow architectures where unaligned
DMA is expensive to redefine NET_IP_ALIGN to 0. But in this case you
are not DMA'ing into the new buffer but using memcpy so you always
want to reserve 2 (16 - ETH_HLEN) to cause aligned access in IP receive.
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
