Hi,
I think this patch could be the final version for now. It has been tested
on two platforms (power and x86_64) and works very well.
Apart from David Miller and Evgeniy Polaykov, we'd like to thank especially
Andrew Gallatin for his great reviews and help to make that happen.
After some discussion we decided to post the LRO patch separately from the
driver patches. Our final driver patches for LRO will be posted later with
some additional fixes for upstream inclusion to the netdev git.
However, I'll also post our LRO patch for the driver today as an example
of how to use this interface.
Thanks a lot,
Jan-Bernd
[PATCH 1/1] lro: Generic Large Receive Offload for TCP traffic
Changes to http://www.spinics.net/lists/netdev/msg37084.html
1) Fixed the LRO_MAX_PG_HLEN bug
2) skb->ip_summed can now be defined by driver for aggregated packets
3) The problem that the "ramp up" for tcp connections between machines
with different MTU size (1500 vs 9000) is very slow has been fixed
by setting skb->gso_size.
4) Checksum problem for little endian machines has been fixed
5) missing additon of vlan_hdr_len for TCP header determination has been added.
From: Jan-Bernd Themann <[email protected]>
Date: Fri, 3 Aug 2007 14:41:14 +0200
> I think this patch could be the final version for now. It has been tested
> on two platforms (power and x86_64) and works very well.
I checked in the LRO patch and the two sample driver ports
to net-2.6.24, thanks!
David Miller wrote:
> From: Jan-Bernd Themann <[email protected]>
> Date: Fri, 3 Aug 2007 14:41:14 +0200
>
>> I think this patch could be the final version for now. It has been tested
>> on two platforms (power and x86_64) and works very well.
>
> I checked in the LRO patch and the two sample driver ports
> to net-2.6.24, thanks!
Good to hear, thanks all.
I'll ponder whether I want to rebase netdev-2.6.git#upstream (2.6.24
queue) on top of net-2.6.24. I might put it off until the pain
threshold rises, or I might go ahead and do it. Undecided.
Either way, I'll want you to push to Linus before I do, when the next
merge window opens.
Jeff
From: Jeff Garzik <[email protected]>
Date: Thu, 09 Aug 2007 02:31:11 -0400
> Either way, I'll want you to push to Linus before I do, when the next
> merge window opens.
No problem.