Subject: Re: sparc64 / 2.0.1 and thread crashes (was: Re: Ultra 5 / 2.0 / panic: lockmgr: no context)
To: Gert Doering <gert@greenie.muc.de>
From: Eduardo Horvath <eeh@NetBSD.org>
List: port-sparc64
Date: 02/08/2005 17:09:25
On Tue, Feb 08, 2005 at 09:37:36AM +0100, Gert Doering wrote:
> Hi,
> 
> to summarize and close down this thread...
> 
>   - my U5 with 2.0.1 kept crashing every night in "low-duty" periods,
>     with varying kernel error messages - sometimes even with RED STATE
>     and SIR Reset messages that pointed to hardware issues

SIR is a software initiated reset.  It's not really what I'd
consider a hardware problem.  A number of SIR instructions are
sprinkled in significant corners of the low-level trap handlers.
They are usually inserted in places where a series of highly
unlikely or impossible events have occurred and the kernel is
unable to clean up the situation.

RED state exceptions occur when the processor runs out of trap 
levels.  This usually means that a trap has occurred inside 
another trap handler.

To get this sort of thing fixed you need to dump the trap frames
from the OBP and look up the corresponding code in the kernel
to find out how you got into that state in the first place.

Most likely this has something to do with being unable to
locate the kernel stack.  But I really shouldn't speculate
without hard data.

Eduardo