Harald,

Thanks a lot for the clarification. If NAT for  non-locally generated
packets means no TSO possible, then we are not including all those
large base of private-addressed TSO-enabled LAN hosts  from taking its
benefits. The NAT box in fact would be handling load from many local
hosts and makes TSO all the more important to maintain performance
levels,  and I hope  we can somehow make it possible.

I have a gut feeling that it is possible to merge  the fast arriving
back-to-back segments from the local  LAN  into corresponding jumbo
frames and submit to the device so send. Too much of hand waving ?
Sorry!

Regards
Anand

On 8/3/05, Harald Welte <[EMAIL PROTECTED]> wrote:
> On Sat, Aug 13, 2005 at 03:20:09PM +0530, Anand SVR wrote:
> > Hi,
> >
> > While browsing through the code, I encountered instances in
> > ipv4/netfilter where checksum related calls are made in ip_nat_core.c
> > and other files. Wondering if netfilter should be made aware of tso so
> > that the checksum operation can be avoided altogether under such
> > circumstances since hardware is anyway going to do it.
> 
> TSO can only happen for locally-generated packets, am I right?  If we
> receive packets from the incoming network, they will be MTU-sized.
> Assuming that the MTU of incoming and outgoing interface are sort-of
> equal, I don't think we would ever see TSO...
> 
> So we're talking about optimizing the case where a user has configured
> NAT for packets that are originated by local sockets?  I personally
> don't think it's worth optimizing that case.
> 
> However, if someone sends me patches (and they look clean and don't
> introduce too much additional complexity), I think I would apply them.
> But it doesn't sound like something that's worth putting time into it..
> 
> --
> - Harald Welte <[EMAIL PROTECTED]>                   http://gnumonks.org/
> ============================================================================
> "Privacy in residential applications is a desirable marketing option."
>                                                   (ETSI EN 300 175-7 Ch. A6)
> 
> 
>
-
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

Reply via email to