[LinuxBIOS] hotplugging flash piece on A8N-E / ROMSIZE

[Philipp Degler]

On Friday 31 August 2007 18:49:55 Robert Millan wrote:
> On Thu, Aug 30, 2007 at 06:57:18PM +0200, Philipp Degler wrote:
> > On Monday 27 August 2007 00:31:37 Quux wrote:
> > > Peter Stuge schrieb:
> > > > It may not be best practice to hotplug, but it is rather safe since
> > > > the chip is never accessed once the system is running.
> > >
> > >  saves plug cycles and time ...
> > >
> > > >> Even an 8 Mbit piece 49LF080A works with the Asus a8ne, so
> > > >> maybe a kernel may be squeezed in there as well.
> > > >
> > > > Yup, but only with a minimum of drivers.
> > >
> > > a 2 MByte piece was used to contain LB + Kernel. I guess a 2 MByte
> > > flash would work in A8N-E , too.
> > >
> > >
> > > currently I am trying to manage with the variables ROMSIZE , IMAGE SIZE
> > > asf. to build a ROM with correct size 512 KB, which seems tricky for
> > > some obscure reason.  --Q
> >
> > Today I checked out the latest svn revision and tested on our ASUS A8N-E
> > rev 2.0.
> >
> > Status of the board is unchanged. It boots with filo 5.0 from IDE. The
> > autmatically build images are not working. Reason is the RAM issue with
> > Athlon64 CPU in 939 socket.
> >
> > You need to modify amdk8/raminit.c
> >
> > Index: src/northbridge/amd/amdk8/raminit.c
> > ===================================================================
> > --- src/northbridge/amd/amdk8/raminit.c (Revision 2754)
> > +++ src/northbridge/amd/amdk8/raminit.c (Arbeitskopie)
> > @@ -1201,7 +1201,7 @@
> >         if (unbuffered && registered) {
> >                 die("Mixed buffered and registered dimms not supported");
> >         }
> > -#if 1
> > +#if 0
> >         // yhlu debug: Athlon64 939 can do dual channel, but it uses
> > unbuffered DIMMs
> >         if (unbuffered && is_opteron(ctrl)) {
> >                 die("Unbuffered Dimms not supported on Opteron");
> >
> > I also need to mention that you have to put the RAM module in a specific
> > bank. I used bank 3 (blue).
>
> Did you try memtest86+ ?  Asides from the die() you commented out, there
> were spurious memory access problems.  

Yes I successfully "memtest"ed before commiting a8n-e target. But maybe this 
was a lucky run. I started memtest from the grub-menu of filo-0.5 payload. 

Running memtest directly as linuxbios payload in earlier tests returned errors 
during the random-pattern-test. 

> In my similar board (A8N5X) those 
> problems disappeared with a patch provided by Rudolf.

I would be happy to test that patch on the A8N-E. 

> See also:
>
>   http://www.mail-archive.com/linuxbios@linuxbios.org/msg10078.html
>
> I don't remember if the patch was made public.  You'd better ask him about
> it :-)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part.
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20070901/096b8a25/attachment.sig>