Re: [PATCH,RFC 2.6.14 11/15] KGDB: ppc64-specific changes

[Tom Rini <trini@kernel.crashing.org>]

On Fri, Nov 11, 2005 at 11:55:22AM +1100, Paul Mackerras wrote:
> Tom Rini writes:
> 
> > This adds basic KGDB support to ppc64, and support for kgdb8250 on the 'Maple'
> > board.  All of this was done by Frank Rowand (who is on vacation right now,
> > but I'll try and answer for him).  This should work on any ppc64 board via
> > kgdboe, so long as there is an eth driver that supports netpoll.  At the
> > moment this is mutually exclusive with XMON.  It is probably possible to allow
> > them to be chained, but that sounds dangerous to me.  This is similar to
> > ppc32, but ppc32 does not explicitly test.
> 
> We already have infrastructure to allow either xmon or kdb to be used,
> and in fact both can be built in and you can select at runtime which
> you prefer.  See the __debugger stuff in system.h.  You should just be
> able to hook kgdb into that same infrastructure.  We're also planning
> to move to using the die_notify stuff for getting all the significant
> events to the debugger.

The notify_die stuff is wonderful.  If I had a ppc64 board, I'd give
that a whirl myself.  This does, unless there was something I missed,
tie into the existing debugger pointer stuff (thats similar but
different from ppc32).  Assuming it hasn't started already, if I can
find time I might give it a wack on arch/powerpc, since I assume my
AlBook works (if not, I see my LongTrail might ;))

-- 
Tom Rini
http://gate.crashing.org/~trini/