[okl4-developer] Roottask pagefault

[Geoffrey Lee]

On Wed, Feb 20, 2008 at 06:37:46PM +1100, Ryan Heffernan wrote:
> 
> > 
> > Hi
> > 
> > I think the problem is that you may be passing in an invalid pd
> > reference, the pd_ref_t appears to be cast to a pd structure pointer
> > in the Iguana server directly then passed to the is_pd() macro
> > to be dereferenced.
> > 
> > 	-gl
> > > 
> > > 
> > > _______________________________________________
> > > Developer mailing list
> > > Developer at okl4.org
> > > https://lists.okl4.org/mailman/listinfo/developer
> > > 
> > 
> Thanks for the reply!
> 
> It does appear that this is the problem (it's a garbage value), but what
> is troubling me is that I can't find anywhere in the code I've added
> that makes the call that provides this pd_ref value to the iguana
> server. However, when I examine the tcb of my server in the gumstix
> kernel debugger, the message is right there in the message register. So
> I guess now my question is if you know of any feature of iguana (or the
> build system) that sends messages to the iguana server on behalf of
> other components (ie. to set up memory, protection domains, etc). 
> 


Hi

Do you mean you also see this garbage value in the message
registers?  Or do you mean you see the correct value in the
register?

There are several things you can check:

(1) check that the message register for the Iguana server and the
application that made the call matches up, because the message
registers are copied in the IPC transfer

(2) If the source message register contains garbage, it might
be useful to trace the call stack to see how it ended up making
that RPC call to the Iguana server in the first place.

(3) If the destination message register is sane, perhaps check
the Magpie code to see if it is unmarshalling the data properly,
and whether there is anything missing between the RPC request 
being received and the is_pd() macro being called.


	-gl

> Ryan 
> 
> 

--