[Avila] Too many ethernet interrupts on Gateworks board

[Burns, Darrell]

You can always do what I did and change the frequency of the interupts
in the IAL.  By adjusting some buffer sizes, we cut the 4000
interrupts/second in half.

Darrell

-----Original Message-----
From: Tim Harvey [mailto:tim_harvey at yahoo.com]=20
Sent: Wednesday, February 28, 2007 4:54 PM
To: Avila
Subject: Re: [Avila] Too many ethernet interrupts on Gateworks board
Subject: [Avila] --- David Acker <dacker at roinet.com> wrote:

> Jay Lee wrote:
> > I have GW2348-4 boards and using gateworks development kit to
compile=20
> > kernel. I'm experiencing exactly same problem of 2000-4000
interrupts=20
> > per second. I've used kernel 2.4 and kernel 2.6 but the result was
same.=20
> > I've also updated madwifi driver to latest and tried on both 2.4 and
2.6=20
> > kernels but still had the same problem.
> >=20
> > I gave up the development software from gateworks and used snapgear=20
> > 3.4.0 with Intel CSR 2.0. With this, I still noticed high interrupts

> > however cpu utilisation was 0. I'm not sure if it has resolved the=20
> > problem or cpu utilisation is being measured correctly. Moreover the

> > default snapgear did not detect the onboard flash properly.
> >=20
> > Can anyone suggest me easy ways to solve this high cpu util / high=20
> > interrupts problem. Is there any snapgear patch to detect the
onboard=20
> > flash?
> >=20
>=20
> The problem is the intel ethernet driver.  It does some kind of
periodic=20
>   interrupt that occurs every often that the driver uses to trigger
some=20
> kind of polling.  I got rid of that driver and went with a real open=20
> source driver, http://www.hohnstaedt.de/ixp_npe/ .  I am running 0.2.1

> with 2.6.18.4, although 0.3.1 went through a review on the ARM list
not=20
> too long ago.  I believe that this driver will go into the mainline in

> the near future.  There may be a way to configure the intel driver to=20
> fix this but I have not tried it.  Personally, I find the hohnstaedt=20
> driver much easier to follow.
> -Ack

I'm not convinced the high interrupt rate with the IAL driver is
necessarily a
'problem'.  I'm running the GPL 0.3.1 driver in a 2.6.19 kernel and have
not
seen any real performance boost over the IAL driver (yet I like it much
better
as it integrated nicely into new kernels and will likely be present in
the next
kernel whereas the IAL will always be a nightmare to patch into a
kernel).

Tim

>=20
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: avila-unsubscribe at lists.unixstudios.net
> For additional commands, e-mail: avila-help at lists.unixstudios.net
>=20
>=20

---------------------------------------------------------------------
To unsubscribe, e-mail: avila-unsubscribe at lists.unixstudios.net
For additional commands, e-mail: avila-help at lists.unixstudios.net