RE: [E1000-devel] Workaround the 'Tx Unit Hang' problem

[Ronciak, J. <joh...@in...>]

Can you get us the syslog (after the hangs have happened) as well.  We
are probably dumping some information there that may help us out with
the hangs.

Thanks.

Cheers,
John


> -----Original Message-----
> From: Dan Aloni [mailto:da...@mo...]=20
> Sent: Tuesday, November 08, 2005 10:01 AM
> To: Ronciak, John
> Cc: Brandeburg, Jesse; e10...@li...
> Subject: Re: [E1000-devel] Workaround the 'Tx Unit Hang' problem
>=20
>=20
> On Tue, Nov 08, 2005 at 09:41:53AM -0800, Ronciak, John wrote:
> > Dan,
> >=20
> > What kernel are you running?  Did turning off TSO using=20
> ethtool work?
> > Since you are running at gigabit speed nothing needs to be=20
> done.  The
> > time difference you are see with the code patch you=20
> supplied below is
> > most probably due to spanning tree on your switch which is=20
> delaying the
> > auto-negotiation.  You can try to disable that on your=20
> switch which will
> > speed linking.
> >=20
> > This really looks like the TSO bugs we have seen in the=20
> past. Please get
> > back to us with the kernel version and disabling TSO results.
>=20
> Hey John.
>=20
> According to ethtool it seems that TSO is not supported by=20
> the card I have:
>=20
> "Cannot get device tcp segmentation offload settings:=20
> Operation not supported"
>=20
> About the kernel version, I am running 2.4.31 + e1000 6.2 +=20
> NAPI enabled.
>=20
> I'd look into disabling the spanning tree stuff in the switch.
>=20
> Thanks,
>=20
> --=20
> Dan Aloni
> da...@mo..., da...@co..., da...@gm...
>=20