Re: mbuf_outbound_finalize bug?
Re: mbuf_outbound_finalize bug?
- Subject: Re: mbuf_outbound_finalize bug?
- From: Andrew Gallatin <email@hidden>
- Date: Wed, 3 Jan 2007 08:50:22 -0500 (EST)
Bhavesh Davda writes:
>
> It was common for older NICs to only support TCP/UDP offload, and not
> IP header checksum offload, because IP header checksum offloads don't
> take much CPU, and probably for some of the hardware
> errata/limitations that Adi mentions below. So I'm not surprised that
> the Yukon NIC doesn't support IP header checksum offload.
>
FWIW, we have a new PCIe 10GbE NIC which supports just TCP/UDP
checksum offload, and not IP header checksum offload (unless you're
using TSO). This is simply because a modern CPU can calculate the IP
header checksum essentially for free (the header should be in L1 cache
when it is checksummed), so there is little point to parsing the
header and doing it in a NIC.
Drew
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden