Re: Crash on Cobalt with CONFIG_SERIO=y

[Peter Horton <pdh@colonel-panic.org>]

On Tue, Feb 28, 2006 at 04:54:05PM +0000, Martin Michlmayr wrote:
> * Peter Horton <pdh@colonel-panic.org> [2006-01-21 01:04]:
> > > >  Real Time Clock Driver v1.12a
> > > >  Cobalt LCD Driver v2.10
> > > >  i8042.c: i8042 controller self test timeout.
> > > >  Unhandled kernel unaligned access[#1]:
> > > 
> > > The i8042 error message is a little surprising.  The Cobalt boards afair
> > No SuperIO, but there is a bog standard VIA PCI-ISA bridge which
> > contains a bog standard PS/2 keyboard controller, which you would have
> > thought should just work ...
> 
> That's really interesting.  I tried latest git (2.6.16-rc5) to see if
> this is still there.  First of all, 32-bit is fine, both regarding
> i8042 as well as ALSA with a PCI audio card.  With a 64-bit kernel,
> i8042 appears to be okay now.  I get:
> 
> Cobalt LCD Driver v2.10
> i8042.c: i8042 controller self test timeout.
> Serial: 8250/16550 driver $Revision: 1.90 $ 4 ports, IRQ sharing disabled
> 
> and the kernel boots fine.
> 
> I then built ALSA support into the kernel to see whether
> http://www.linux-mips.org/archives/linux-mips/2006-01/msg00325.html
> disappeared too.  I got almost the same oops as the one I got before
> because of i8042.
> 
> Peter, does that make any sense to you?  I assume your recent PCI fix
> may be related.
> 

Those addresses that begin 9b640000 (including the fault address) look
very strange. The low 32-bits look like a valid physical address in the
PCI space but the top bits definitely don't look right (unless the
kernel's playing tricks with unused address bits; I'll have to check the
RM523x data sheet to see if they have any effect). Have any of the MIPs
experts commented ?

> 
> BadVA : 9b64000001014c87
>

P.