[EMAIL PROTECTED] wrote:
> Bill Sommerfeld wrote:
>
>> On Thu, 2007-09-13 at 12:01 -0400, Sebastien Roy wrote:
>>
>>
>>>> I still believe that a much better solution would be to offer
>>>> drivers a way to "Punt" on a packet.
>>
>> Absolutely. Define a function they can call to finish the checksum in
>> the event they have a hardware limitation that prevents the use of
>> checksum offload on a particular packet.
>>
>>
>
> On a vaguely related note, it might be an interesting exercise
> to make IP assume that *all* NICs provide checksum offload
> support and to put the checksum'ing of IP packets into the
> mac/dls layer.
A few possible issues:
1) Not all NICs are GLDv3. UV could address this I suppose
2) What about IP configurations where the IP checksum is
"embedded"... e.g. IPsec, IP-in-IP tunnels, and so forth?
Were it not for the complexities of tunnels, it probably would be a good
idea, because it would simplify IP. But alas, I don't think you can
make those assumptions.
-- Garrett
>
> Darren
>
_______________________________________________
networking-discuss mailing list
[email protected]
