[LinuxBIOS] m57sli-s4 build tutorial

[Ward Vandewege]

I've put up a build tutorial for the m57sli-s4 at

http://linuxbios.org/M57SLI-S4_Build_Tutorial

It's still based off of YH's big patch, but if Ed's patch from today gets
merged, I will update the tutorial to use the v2 tree.

Remaining issues are ACPI support, and the mac address changes (but Stefan
might have a way to fix that).

Thanks,
Ward.

[Stefan Reinauer]

* Ward Vandewege  [070319 23:03]:
> I've put up a build tutorial for the m57sli-s4 at
>
> http://linuxbios.org/M57SLI-S4_Build_Tutorial
>
> It's still based off of YH's big patch, but if Ed's patch from today gets
> merged, I will update the tutorial to use the v2 tree.
>
> Remaining issues are ACPI support, and the mac address changes (but Stefan
> might have a way to fix that).

layouts are not good for exclusion, so it would be the easiest to say

(for a 512k image)

flashrom -w -s 0x3ffd0 -e 0x3ffdf image.rom

This is only for the case that a linuxbios image was previously flashed.

[Peter Stuge]

On Mon, Mar 19, 2007 at 11:14:22PM +0100, Stefan Reinauer wrote:
> Ward Vandewege  [070319 23:03]:
> > Remaining issues are ACPI support, and the mac address changes
> > (but Stefan might have a way to fix that).
>
> layouts are not good for exclusion, so it would be the easiest to
> say
>
> (for a 512k image)
>
> flashrom -w -s 0x3ffd0 -e 0x3ffdf image.rom
>
> This is only for the case that a linuxbios image was previously
> flashed.

Here image.rom is the factory BIOS, correct?

//Peter

[Stefan Reinauer]

* Peter Stuge  [070319 23:51]:
> > flashrom -w -s 0x3ffd0 -e 0x3ffdf image.rom
> >
> > This is only for the case that a linuxbios image was previously
> > flashed.
>
> Here image.rom is the factory BIOS, correct?

No, LinuxBIOS.

it just ignores the 16 bytes at ffffffd0

[Peter Stuge]

On Mon, Mar 19, 2007 at 11:30:19PM +0100, Stefan Reinauer wrote:
> Peter Stuge  [070319 23:51]:
> > > flashrom -w -s 0x3ffd0 -e 0x3ffdf image.rom
> > >
> > > This is only for the case that a linuxbios image was previously
> > > flashed.
> >
> > Here image.rom is the factory BIOS, correct?
>
> No, LinuxBIOS.
>
> it just ignores the 16 bytes at ffffffd0

Ok, sorry if I'm daft, but how does the MAC address end up in the
LinuxBIOS rom file?

Or, aha, did you mean "This is only for the case that a factory image
was previously flashed." instead of the other way around?

//Peter

[Stefan Reinauer]

* Peter Stuge  [070319 23:59]:
> > it just ignores the 16 bytes at ffffffd0
>
> Ok, sorry if I'm daft, but how does the MAC address end up in the
> LinuxBIOS rom file?
> Or, aha, did you mean "This is only for the case that a factory image
> was previously flashed." instead of the other way around?

No, it does not help that problem. Above line only helps when updating
from an old working LinuxBIOS to a new one.

For getting the mac address from the machine's legacy bios, we would
have to add a whole bunch of logic to flashrom:

* Determine that it's a transition from awardbios to linuxbios (safely)
* read the mac address (will it always be the same position for all
  award bios updates in the future?)
* patch the to-be-flashed linuxbios image on the fly.

Such a transition is not exactly free of risks.

There might be another solution if we accept that the machine's mac
address will change.

We could use 4 fixed bytes and generate the last 4 bytes from
/dev/urandom during _build_. This would leave us with a working machine
with a reasonable mac address. Then, with above line you prevent that
the mac address changes between linuxbios updates.

Stefan

[Lu, Yinghai]

Let the flashrom take extra command parameter. MB should have MAC
address stick on the side.

YH

[Ward Vandewege]

On Mon, Mar 19, 2007 at 04:02:19PM -0700, Lu, Yinghai wrote:
> Let the flashrom take extra command parameter. MB should have MAC
> address stick on the side.

Or it can be easily determined with ifconfig.

Yes, that's a great idea.

Thanks,
Ward.

[Peter Stuge]

On Mon, Mar 19, 2007 at 04:02:19PM -0700, Lu, Yinghai wrote:
> Let the flashrom take extra command parameter. MB should have MAC
> address stick on the side.

Either that, or add a Config.lb option for the MAC since it will
always be the same for each board. Plus it only makes sense for
some boards.

//Peter

[Lu, Yinghai]

You will build rom for each MB if you have 10 or 100 same MB?

YH

-----Original Message-----
From: [e-mail]
[mailto:[e-mail]] On Behalf Of Peter Stuge
Sent: Monday, March 19, 2007 5:28 PM
To: [e-mail]
Subject: Re: [LinuxBIOS] m57sli-s4 build tutorial

On Mon, Mar 19, 2007 at 04:02:19PM -0700, Lu, Yinghai wrote:
> Let the flashrom take extra command parameter. MB should have MAC
> address stick on the side.

Either that, or add a Config.lb option for the MAC since it will
always be the same for each board. Plus it only makes sense for
some boards.

//Peter

[Stefan Reinauer]

* Peter Stuge  [070320 01:28]:
> On Mon, Mar 19, 2007 at 04:02:19PM -0700, Lu, Yinghai wrote:
> > Let the flashrom take extra command parameter. MB should have MAC
> > address stick on the side.
>
> Either that, or add a Config.lb option for the MAC since it will
> always be the same for each board. Plus it only makes sense for
> some boards.

A combination of both would be nice. the config option
MAC=`/sbin/ifconfig eth0|grep HWaddr|cut -f2 -dH|cut -f2 -d\ `
flashrom -mac $MAC ....

If it is not specified, it would use the value from the config option.

Can someone make a patch?

[Ward Vandewege]

On Mon, Mar 19, 2007 at 11:58:09PM +0100, Stefan Reinauer wrote:
> Determine that it's a transition from awardbios to linuxbios (safely)
> read the mac address (will it always be the same position for all
> award bios updates in the future?)

Of course there is a bit more information that can be used here - the factory
mac address will be somewhat predictable, if we know the board we're
flashing. Only the last 6 of the 12 bytes of the mac address are an unknown.
In my case, the 00:16:e6 is Gigabyte's vendor id, unsurprisingly
(http://www.coffer.com/mac_find/?string=00%3A16%3Ae6).

Thanks,
Ward.