Re: using kernels bigger than 1966080 bytes on the nslug with slugimage
On Wed, Feb 11, 2009 at 02:36:20PM +0100, Loïc Minier wrote:
> On Tue, Feb 10, 2009, Marc Singer wrote:
> > I responded to L. Minier with the suggestion that you move the load
> > address for the kernel to something after APEX. Try 0x01000000. You
> > could also change the VMA address for APEX to be 3MiB from the base of
> > RAM instead of 2MiB.
>
> It was an excellent suggestion, it worked just fine; thanks!
>
> I'm attaching your message here (I hope that's ok with you, don't think
> anything was private) since it has helpful information and Oliver
> brought up the problem here as well, so I can as well attach the
> solution.
>
> I think Oliver tried our various ramdisk sizes and started seeing
> issues with ramdisks of size 5636080 and above; I didn't research what
> these issues are as I don't have access to a NSLU2 myself, but I guess
> it might be interesting to someone to look into making the APEX default
> config accept any theoritical kernel and initrd size in the 0 - 8M
> range (since that's the size of the flash).
>
> Oliver, if you think the APEX changes wouldn't harm on Debian could you
> please share them here as to allow Marc (who's also Debian maintainer
> of APEX I believe) to perhaps merge them? I don't think official
> Debian builds need them right now, but it could help users building
> custom larger kernels or in the future if the kernel size grows as
> Ubuntu's did.
The default load addresses for the kernel and ramdisk are 0x00008000
and 0x01000000, but these are part of the target-specific
configuration. The kernel load address is probably OK if we move APEX
a little further out in RAM. The ramdisk load address should, in of
itself, be OK. If you are having problems with larger ramdisks then
you may be running into a limit in the kernel.
Reply to: