On Wed, May 23, 2007 at 17:07:17 +0200, Quentin Garnier wrote:

> On Wed, May 23, 2007 at 10:32:56AM -0400, Blair Sadewitz wrote:
> > I was looking at this today and was wondering if this could work on
> > i386.  My motherboard (Intel D945G) takes an inordinate amount of time
> > to reboot sometimes, and I reboot often enough such that this would be
> > convenient.
> 
> kloader(9) currently cannot work on archs where you don't have a memory
> space where va == pa is true.
> 
> So, kloader(9) on i386 is possible barring that you:
> 
>   1.  change the kloader inner workings lightly so it copies the new
>       kernel into a set of regular virtual addresses.
>   2.  create an api (in uvm or through some hack of your choice, it can
>       be made better later) to easily get the few pages you need for the
>       bootloader that verifies pa == va.
>   3.  pass the set of physical addresses where the kernel has been
>       copied to said bootloader.
> 
> At least, that's my recollection of when I looked at it for hpcarm.

Check this thread on port-arm:

    http://mail-index.netbsd.org/port-arm/2006/12/04/0001.html

that might be an interesting alternative to kloader.

SY, Uwe
-- 
uwe@stderr.spb.ru                       |       Zu Grunde kommen
http://snark.ptc.spbu.ru/~uwe/          |       Ist zu Grunde gehen