On Wed, Feb 07, 2001 at 07:02:18PM -0700, Michael L. Hitch wrote:
> On Wed, 7 Feb 2001, Mark Abene wrote:
> 
> > Unfortunately since "trace/u" is not implemented in mips's ddb, it's making
> > this init problem hard to debug.  All I have to work with is the following:
> > 
> > PC 0x400180: not in kernel space
> > 0+400180 (400180,c4705f68,0,0) ra 0 sz 0
> > User-level: pid 1
> > 
> > Could someone with a working NetBSD-1.5/arc install tell me where in init
> > that address is?  It's always sitting at that address...
> 
>   If the arc binaries are the same as pmax, that address is the starting
> address of init.  That makes it sound like maybe the process never
> actually gets started.
> 

Makes sense.  I'm guessing that init is always forking and returning.
Sometimes the execve'd process begins to start, which is why I sometimes
see a "pid 10" called init (and once called "sh").  Other times the execve'd
process never properly starts, so it's absent from the process table.
In any event, pid 1 is going through its normal event loop so I see it
at its start location whenever I break to DDB.  Just a theory.  As to why
init is having trouble forking, is anyone's guess.
I'm hoping more people will try my code on their Magnums and post their
results!

Cheers,
-Mark