bug-grub
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: problem with grub configuration


From: Jeremy Katz
Subject: Re: problem with grub configuration
Date: 08 Feb 2002 16:01:51 -0500

On Fri, 2002-02-08 at 15:50, address@hidden wrote:
> Jeremy Katz <address@hidden> wrote:
> 
> > > Do this with the "uppermem" command, so instead of your entry looking
> > > like the following:
> > [snip] 
> > > ...so I *think* it would instead look like this:
> > > 
> > > title Red Hat Linux (2.4.7-10smp)
> > >         uppermem 259072
> > >         root (hd0,0)
> > >         kernel /boot/vmlinuz-2.4.7-10smp ro root=/dev/sda1
> > >         initrd /boot/initrd-2.4.7-10smp.img
> > 
> > You also still need to pass mem= on the kernel command line if you go
> > this route.  Having grub pass its detected memory to the kernel by
> > default with a 2.4 kernel leads to problems due to not taking into
> > account any holes which you may have in your memory layout, thus is
> > disabled in the Red Hat Linux packages.
> 
> What?  OK, I'm confused.  Do you mean, removed from the Red Hat Linux
> *GRUB* package?

We run configure with --disable-auto-linux-mem-opt (as do most distros
shipping 2.4 kernels from my looking through others' packages).

> I looked at the Red Hat kernel source, and it seemed clear that the
> kernel already handles this just fine.
> 
> It parses the "mem=" option as just one memory region.  I even tested
> this out on one of my huge memory boards recently (from recent GRUB
> CVS), and it worked perfectly fine, restricting one memory region
> while allowing for the others.
> 
> GRUB's "mem=" option it passes takes into account memory holes just
> fine, but not putting anything above the first hole into there, which
> is what the kernel source expects.
>
> Maybe I'm missing something here?

>From what I understand from talking with our kernel folks, using mem=
without passing along the complete e820 memory map info is not
guaranteed to give proper memory information in all cases.  Thus,
passing mem= by default without an e820 map can cause problems.

But, IANAKH (I am not a kernel hacker :) and instead just listen to what
those who in theory know more about these things give me information :-)

Cheers,

Jeremy



reply via email to

[Prev in Thread] Current Thread [Next in Thread]