> From: David Brownlee <abs@netbsd.org>
>
> On Tue, 15 Feb 2000, Ross Harvey wrote:
>
> > > From: David Brownlee <abs@netbsd.org>
> > >
> > > 	Would it be feasible to build a special multa bootloader that
> > > 	used the old buggy milo PALcode to load the kernel? Once was
> > > 	loaded everything could switch to SRM?
> > 
> > Uhh, if you had SRM, you wouldn't be booting from ARC. If you were booting
> > from ARC, there would be no SRM to switch to. And you boot from ARC
> > using the NT palcode, by making ARC callbacks...the thing you need the
> > OSF/1 palcode for is running the kernel, not for booting it.
> > 
> > Or did I completely misunderstand the question?
>
> 	Could you embed the Multia SRM in a bootloader, similar to
> 	milo?


Well, that's an interesting idea. I presume you meant (no comma) "a bootloader
similar to milo".

My comments:
	
	* Embedding the entire SRM image is really cute, I never thought
	  of that...except...if you have SRM, why not just flash it? And
	  it probably decompresses itself out of flash .. somehow I think
	  the amount of reverse engineering and binary patching to make it
	  handoff from the SROM to your code to SRM would be far better
	  spent elsewhere on NetBSD.

	* Doing that might have some copyright and distribution hangups.
	  DEC^WCompaq is very protective of SRM, as it essentially reserves
	  OSF/1^WTru64 and VMS for their own HW.

	* The palcode bugs only apply to ev56. (Well, maybe ev5 also.)
	  I'm sure the milo palcode is fixable, or that a kernel workaround
	  could be used. I will say, though, that I tried with no success
	  for a few days to fix the ebsdk palcode once, before dec stepped
	  in and helped out the project I was on with the real palcode.

Ross