Re: ack! grub bug...

[Jeff Garzik]

Jochen Hoenicke wrote:
> 
> On Feb 23, OKUJI Yoshinori wrote:
> > From: Jeff Garzik <address@hidden>
> > Subject: Re: ack! grub bug...
> > Date: Thu, 22 Feb 2001 18:57:47 -0500
> >
> > > hrm, well... the thing is, hardware is out there, and more is coming
> > > out, where mem=XX simply will break on that machine.  In the case of a
> > > broken bios (bad e820, etc.), the boot loader would need to pass memory
> > > regions to the kernel, so IMHO having the rule "always pass mem=XX by
> > > default" breaks on newer hardware.
> >
> > So I said that that should be fixed in Linux. I don't change anything,
> > unless you give me a good explanation of why that shouldn't be
> > addressed by Linux.
> 
> If I understand correctly, the problem with mem=XX and linux-2.4 is
> when there are memory holes.  It is quite clear that old Linux
> versions have the same problem with memory holes, since there is no
> way to tell the kernel where they are.  So it wouldn't break anything
> if grub would omit the mem=XX parameter when it detects a memory hole.

This will solve the problem... for problematic machines.  However, it is
merely a bandaid... policy is that mem=XX is an -override- parameter, to
be used only when necessary.  IMHO we should move back towards
defaulting grub to --no-mem-option in general, in the future. 
(apparently, it cannot be done now, for the general grub release)

As for now, I think MandrakeSoft will make a vendor patch that defaults
to --no-mem-option, since we can control our kernel versions a bit
more...

        Jeff


-- 
Jeff Garzik       | "You see, in this world there's two kinds of
Building 1024     |  people, my friend: Those with loaded guns
MandrakeSoft      |  and those who dig. You dig."  --Blondie