Garrett D'Amore wrote:
> On Tue, 2007-07-31 at 17:51 +0200, Darren Reed wrote:
>   
>> Garrett D'Amore wrote:
>>     
>>> ...
>>> I _think_ it works out that for receive there is no problem... the
>>> checksum will be verified fine ... but unfortunately we have no way in
>>> the stack to register to provide rx checksum offload without also
>>> offering to provide tx checksum offload.
>>>       
>> Heh...and on BSD this would be done with:
>>
>> # ifconfig hme0 udp4csum-rx -udp4csum-tx
>>
>>
>> In light of "this problem", it might be worth looking at extending
>> Brussels to enable Solaris to provide the same level of management
>> for "the future"....
>>
>>
>> Has anyone verified whether or not the same problem exists with
>> another operating system such as NetBSD (I don't have any handy
>> with hme's in them) to get an idea about whether it is the OS or
>> the hardware?
>>     
>
> It would appear that NetBSD, at least, suffers from the same problem
> with TCP and UDP on hme interfaces.
>
> Interestingly enough, I'm not confident that NetBSD won't have the same
> problem with tiny packets that we enountered.  (Much of the original hw
> cksum implementation I borrowed from NetBSD originally.... to avoid
> taining with qfe sources which are not yet open.)
>   

Yes, it has been verified with small UDP packets on NetBSD.
:-(

Darren

_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to