[FreeBSD 8.2 amd64 XENHVM] Incorrect behavior of xen networking drivers: unexpected tcp reassembling

[laurent.cligny at gmail.com]

Le 12/01/2012 19:42, Ivan Ivanyuk a écrit :
> On 12 January 2012 00:35, Ivan Ivanyuk <ivan.ivanyuk at gmail.com> wrote:
>> Hi All,
Hi Ivan,
>> I can get about 500Mb/s from virtual machine to external host, but
>> only 200Kb/s from any internal PC to the same external host through
>> virtual machine router.
>>
>> Closest description I found in archives is this:
>> http://lists.freebsd.org/pipermail/freebsd-xen/2011-May/000902.html.
>>
I'm the one who had that problem and tried to explain it. I must say
that you explained it much better, so hopefully this issue will be
understandable to everyone ;)
>  Somewhat successful follow-up:
>
> There is still some occurrences of reordered packets but original
> issue was resolved.
>
>  Turned out this mechanism can be turned off by "ifconfig -lro" in
> FreeBSD itself.  I'm not sure if turning LRO on virtual driver by
> default is good idea at all, but obviously LRO implementation for xn
> device doesn't consider DF flag implication on resulting packets.
>
>  I'm not familiar with other LRO algorithms which used in physical
> NIC's. Can someone comment if such problem is normal for it?
>
>  And what is your opinion on using tso, rxcsum and txcsum offloads in
> virtual envirnoment?
>
Glad to know you almost addressed the issue with custom settings. I
think I would try again any time soon, to see if I can work with it in
production environment.

Because of this issue I had to run a Linux/debian vhost in place of
FreeBSD, and I would know if anyone had bandwidth problems with this
kind of setup under the brand new FreeBSD 9.0 ?

I would like to be able to use PF and ALTQ at my gateway for a long
time, and this issue was the only show-stopper I felt into.

Regards,

Laurent Cligny