Re: CVS commit: src/sys/arch

[Andrew Doran <ad%netbsd.org@localhost>]

On Sun, Jan 20, 2008 at 07:01:31PM +0200, Elad Efrat wrote:
> Joerg Sonnenberger wrote:
> >On Sat, Jan 19, 2008 at 11:47:32AM -0500, Perry E. Metzger wrote:
> >>>I'm not saying it wouldn't be useful to have, but it probably needs a
> >>>rethink.  I suspect most people use ddb over a serial port for
> >>>debugging.
> >>There is no serial port on the majority of new x86 machines. There has
> >>been a lot of trouble in debugging ACPI support because most laptops
> >>have (effectively) no remote debugging facility at all.
> >
> >Please don't use ACPI as justification when it mostly doesn't apply.
> >The problems for ACPI debugging were not lack of serial console. A
> >serial console doesn't help you that much right after resume as it is
> >dead as well.
> >
> >Many laptops actually have a far superior interface for debugging:
> >firewire. The support for that is at least seriously underdocumented.
> >Not sure if the gdb/kgdb support exists at all.
> 
> Interesting. In light of the above, Chris's notes, and my own experience
> that the ipkdb code didn't even compile for at least a year or so -- how
> desired is it to get this whole thing working? is it even a realistic
> goal...?

It should be easy to get kgdb and firewire working. Assuming the fwmem
driver works then all that should be needed are some mods to libkvm.
See FreeBSD's libkvm. They have some sort of kd->rawdump boolean but 
that should be folded into the ALIVE macros I think.

Andrew