Nvidia SATA hangups with newer FC5 test kernel

Jonathan Berry berryja at gmail.com
Thu Mar 16 19:48:22 UTC 2006


On 3/16/06, Calvin Dodge <caldodge at gmail.com> wrote:
> On 3/16/06, Jonathan Berry <berryja at gmail.com> wrote:
> > Works fine in 64-bit ; ).  For me at least.  It's not the exact same
> > motherboard, but I have an nForce4 chipset.  What problems did you
> > have with 64-bit FC4?
>
> First, an Emachines system with the ATI chipset (this machine is some
> distance away today, so I can't give you more info).
> The clock ran at double the normal rate. Turning off APIC solved that
> problem, but not the random lockups.

Gag.  ATI X200 chipset or similar?  That one seems to be pervasive in
the name-brand sector for some reason.  Yes, I have seen those
complaints before.  I think this is a hardware issue.  Or at least an
issue of the software and hardware not agreeing on something.

> Second, an Asus K8N-E homebrew system (replaced dead Athlon Socket A
> board).  This one would run fine for a few days, then lockup with
> messages that _seemed_ to be related to i2c activity.
>
> That system DID use one ATrpms module - the stock sk98lin module
> didn't work with the PCI-X NIC we installed. So it's possible the
> fault was with that module.

PCI-X?  Then this is a server class board and different from my
desktop class nForce4.  I guess I can't help you much, then.

> I tried disabling lm_sensors, turning off APIC, and disabling "Cool 'n
> Quiet" in the BIOS. Nothing worked.
>
> In desperation I finally saved the config info, then installed 32-bit
> FC4. The system has been stable ever since.
>
> I know people have been successfully running Opterons for some time.
> My suspicion is that not all the kinks have been worked out of the
> 64-bit kernel (or, perhaps, chipset drivers), and those kinks pop up
> on certain hardware.
>
> I'm not a kernel progammer, so take my suspicion with a bag of salt.

There's no telling, I guess.  There is likely an issue in some driver
somewhere.  It's mostly the same code, as far as I know.  It seems
software will never be perfect, and when it comes to interacting with
hardware, things get even more interesting.

Jonathan




More information about the fedora-test-list mailing list