On Sun, Mar 16, 2003 at 11:15:31AM +0000, Andrew Doran wrote:
> Hi,
> 
> On Sat, Mar 15, 2003 at 10:40:13PM +0000, Matt Dainty wrote:
> 
> > scsibus1 at asc1: 8 targets, 8 luns per target
> > pxg0 at tc0 slot 4 offset 0x0
> > unexpected machine check:
> > 
> >     mces    = 0x1
> >     vector  = 0x660
> >     param   = 0xfffffc0000108060
> >     pc      = 0xfffffc000043c57c
> >     ra      = 0xfffffc000043c49c
> >     code    = 0x100000015
> >     curproc = 0xfffffc00004c94d0
> >         pid = 0, comm = swapper
> > 
> > panic: machine check
> > Stopped in pid 0 (swapper) at   cpu_Debugger+0x4:       ret     zero,(ra)
> > db>
> > 
> > Anyone got any ideas? Anyone built and running with one of these?
> 
> The PMAG-C and PMAG-E are known to work well. I'd suggest putting some
> printf()s in pxg_init(), stic_init() and stic_reset() and other called
> functions to see whats going wrong. See sys/dev/tc/pxg.c and stic.c

Hi Andrew,

I've found the problem, in stic_reset():

---8<---
/* Second PixelStamp */
if (yconfig > 0) {
	si->si_stamp[0x100b0] = config | 8;
	si->si_stamp[0x100b4] = 0;
}
---8<---

Commenting this small block out stops the machine check, and the
hardware appears to work okay. I've now learnt not to trust what I see
on the monitor with the S3 switch flipped to serial. Console seems to
work okay, and colours all appear as they should. I'm tempted to see how
X behaves now...

Does this code error make any sense to you?

Matt