[Devel] Re: [PATCHSET] 2.6.20-lxc8

Benjamin Thery benjamin.thery at bull.net
Wed Mar 28 05:30:26 PDT 2007


Eric W. Biederman wrote:
> Kirill Korotaev <dev at sw.ru> writes:
> 
>> Benjamin,
>>
>> checksumming can be optimized out as well.
>> We had an experimental patch for OpenVZ venet device, which adds
>> NETIF_F_LLTX | NETIF_F_HW_CSUM | NETIF_F_SG | NETIF_F_HIGHDMA
>> features to venet device and avoids additional checksumming where possible
>> (moving RX/TX checksum calculation to hardware).
>>
>> So I guess this is doable in future as well.
> 
> I think I have the checksum bits settable in software with etun already.  If not
> it shouldn't be to hard to add.

I tried to activate the checksum offload on etun to see if it improves 
things, but, unfortunately once activated all my traffic was lost or 
blocked. I didn't spend a lot of time on the issue. May be I'll give 
it another try.

BTW, there is a small bug in etun_set_tx_csum. We can't disable the 
checksum offloading once it has been set. I think it should look like 
this: (sorry I haven't a patch ready)

static int etun_set_tx_csum(struct net_device *dev, u32 data)
{
	if (data)
		dev->features |= NETIF_F_NO_CSUM;
	else
		dev->features &= ~NETIF_F_NO_CSUM;
	return 0;
}


> 
> I don't default to that because depending on your configuration it might not
> be safe.  In particular I think when you are using ethernet bridging we
> need to do the packet checksum immediately off the wire.
> 
> Eric
> 


-- 
B e n j a m i n   T h e r y  - BULL/DT/Open Software R&D

    http://www.bull.com



More information about the Containers mailing list