[mythtv-users] Re: Building a myth box, wanna look over my components?

Mike Isely isely at isely.net
Wed Dec 15 20:41:58 UTC 2004


On Wed, 15 Dec 2004, Yan-Fa Li wrote:

> Mike Isely wrote:
> > On Wed, 15 Dec 2004, Gordon Rimac wrote:
> >
> >
> >>On Wed, 15 Dec 2004 01:21:36 -0600, Derek Gathright <drgath at gmail.com> wrote:
> >>
> >>>NForce2 is bad?  Wow, really?  That what I've had others suggest to
> >>>me.  Anyone else wanna chime in?
> >>>
> >>
> >>I had bad luck with an Abit NF7 (nForce2) board.  I was using it on my
> >>backend and had system freezes frequently (hard lock).  It ended up
> >>being the PVR-250 cards and/or the ivtv drivers.  Once I removed them
> >>the NF7 has been rock solid as a front-end and the tuner cards have
> >>been working fine in an ECS motherboard (SIS chipset) on my backend.
> >
> >
> > You don't find it odd that the nForce2 rig had problems with that
> > hardware?  This is actually an interesting datapoint.  One of my more
> > recent unexplained hard lockups happened while testing a WinTV-Go card in
> > my nForce2 system.  About ten seconds into recording a show the system
> > locked up.  Maybe PCI access has something to do with the problems.  The
> > CD writer-related lockups I've had would fit this because the IDE
> > controller in that case is also a PCI card.
>
> It's not that interesting because there was a well known problem with
> NForce2 chipsets locking up under moderate IO.  This has been fixed in
> all recent kernels.  There was some fixup required in their IO-APIC
> hardware that could be worked around.  I'm running both 2.6.9 and 2.4.27
> on NForce2 hardware with great success.  Infact my mythbackend has
> regular uptimes over 30 days.
>
> Try newer kernels, the ones that shipped originally with FC were
> susceptible to this problem.
>
> http://atlas.et.tudelft.nl/verwei90/nforce2/
>

Yes, I'm familiar with what's on that page.  It appears to be the same
page in the URL I pointed to in a post yesterday.  It doesn't have any
information suggesting that the kernel has finally had fixes applied to
it.

I _am_ running stock 2.6.8.1 now, though admittedly that system hasn't had
a lot of time running 2.6.9 (last time I tried that kernel CD writing was
hosed up for me).  I run the Debian distribution, not Fedora, and I build
kernels from the stock sources with a few choice patches (e.g. new tuners,
the -ac patch series, latest v4l).  The lockups I related above happened
under stock 2.6.8.1 (no patches).

When I first encountered that web page describing these problems I
examined the kernel sources I was working with at the time (July 2004
timeframe and I believe 2.6.7) and concluded that the fixes described on
that page had not made it into the kernel sources yet.  However I will
take a second look with the 2.6.9 kernel.  Your post here is the first
time I've heard specifically that nforce2 fixes might have actually made
it into the stock kernel.

  -Mike

-- 
                        |         Mike Isely          |     PGP fingerprint
    POSITIVELY NO       |                             | 03 54 43 4D 75 E5 CC 92
 UNSOLICITED JUNK MAIL! |   isely @ pobox (dot) com   | 71 16 01 E2 B5 F5 C1 E8
                        |                             |


More information about the mythtv-users mailing list