On Fri, Feb 01, 2002 at 11:02:52AM +0000, Richard Earnshaw wrote:

 > Ok, so it's been a while since I tried to build a new kernel for my shark, 
 > so I'm not entirely sure when this problem might have started to happen.

Sigh, me either.  I've been doing most of my stuff on XScale.

 > Shark kernels (arm32/conf/SHARK) are no-longer booting.  They get as far 
 > as the probe for the FPA (they seem to do this even if FPA support is not 
 > requested) and then hang when taking an undefined instruction trap.

I looked at the problem just now...  it's not just the undefined insn
trap that loses.  Looks like all traps are losing.

Off the top of my head, I'm thinking it could be:

	* Cache stuff (not properly cleaning the vector page?)
	* Vector page re-org

CATS kernels are working, though, right?

Anyway, I'll poke around at it tonight and see if I can't track it down.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>