Hello, One of the possible policies for skb->ip_summed is CHECKSUM_HW.
In this case the device can perform checksums in hardware. According to Linux Device Drivers 3, an example of a such network card is the SPARC HME interface. I looked a bit at the code (it's in /drivers/net/sunhme.c) I want to understans something: When I transmit a packet with a card like this, which support calculating checksome in hardware, I do it by happy_meal_start_xmit() method of the network card (because the hard_start_xmit() method of the device is set to happy_meal_start_xmit() in the init method of the card ). But until it gets to there , it passes many stages in the kernel net stack: in dev_queue_xmit() of dev.c, in ip_finish_output,in ip_finish_output, in ip_send()... How does it know NOT to perform such a calculation of checksum in that long way ? such a calculation will be not redunsdant if I understand it right. And in case it does such an unneedded calculation of checksum - is there a way to prevent it so it will perform it only in the network card level , in HW? TIA, John - 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
