Re: [coreboot] r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Arne Georg Gleditsch]

[e-mail] writes:
> Author: stepan
> Date: 2008-08-20 11:17:30 +0200 (Wed, 20 Aug 2008)
> New Revision: 3529
>
> Modified:
>    trunk/coreboot-v2/targets/tyan/s2912_fam10/Config-abuild.lb
> Log:
> this port seems somehow broken.. Now, is it using FAILOVER, or is it not?!

Hmm.  This was supposed to be more or less equivalent to the approach
taken by the serengeti_cheetah_fam10 target, but I see it isn't quite...

It works as it is (was), but I can have a look at it again if there are
stylistic issues.  I might not have exercised the abuild file, though --
what would be the recommended way of doing that so as to ensure that is
in correspondence with the regular config file?

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Marc Jones]

Arne Georg Gleditsch wrote:
> [e-mail] writes:
> > Author: stepan
> > Date: 2008-08-20 11:17:30 +0200 (Wed, 20 Aug 2008)
> > New Revision: 3529
> >
> > Modified:
> >    trunk/coreboot-v2/targets/tyan/s2912_fam10/Config-abuild.lb
> > Log:
> > this port seems somehow broken.. Now, is it using FAILOVER, or is it not?!
>
> Hmm.  This was supposed to be more or less equivalent to the approach
> taken by the serengeti_cheetah_fam10 target, but I see it isn't quite...
>
> It works as it is (was), but I can have a look at it again if there are
> stylistic issues.  I might not have exercised the abuild file, though --
> what would be the recommended way of doing that so as to ensure that is
> in correspondence with the regular config file?

This was my fault. I missed reviewing the abuild config.

This is now the same as serengeti_cheetah_fam10.

A little background information. The fam10 initialization is a bit 
larger than the K8 code. To make more room in the same 512K ROM that 
most platforms have the normal image was dropped. The other change is 
the use of failover. This makes the XIP CAR code live in only one place 
instead of in the fallback and normal images.

Arne, please test a s2912_fam10 abuild and ack this.

Marc

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Arne Georg Gleditsch]

Marc Jones  writes:
> This was my fault. I missed reviewing the abuild config.
>
> This is now the same as serengeti_cheetah_fam10.
>
> A little background information. The fam10 initialization is a bit
> larger than the K8 code. To make more room in the same 512K ROM that
> most platforms have the normal image was dropped. The other change is
> the use of failover. This makes the XIP CAR code live in only one
> place instead of in the fallback and normal images.
>
> Arne, please test a s2912_fam10 abuild and ack this.

I'll do so.  However, once I start digging into this -- would it make
sense to fix it properly so that the target builds a full 1024K ROM (as
the board is equipped with) and actually includes both the normal and
failover image?

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Stefan Reinauer]

Arne Georg Gleditsch wrote:
> Marc Jones  writes:
> > This was my fault. I missed reviewing the abuild config.
> >
> > This is now the same as serengeti_cheetah_fam10.
> >
> > A little background information. The fam10 initialization is a bit
> > larger than the K8 code. To make more room in the same 512K ROM that
> > most platforms have the normal image was dropped. The other change is
> > the use of failover. This makes the XIP CAR code live in only one
> > place instead of in the fallback and normal images.
> >
> > Arne, please test a s2912_fam10 abuild and ack this.
>
> I'll do so.  However, once I start digging into this -- would it make
> sense to fix it properly so that the target builds a full 1024K ROM (as
> the board is equipped with) and actually includes both the normal and
> failover image?

The image resulting from abuild should work on real hardware.  If the
board comes with 1024K rom, choosing that size is a good idea.

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Arne Georg Gleditsch]

Stefan Reinauer  writes:
> The image resulting from abuild should work on real hardware.  If the
> board comes with 1024K rom, choosing that size is a good idea.

I'll do that, then.  (I am using the 512K image on real hardware today,
but I'm padding it with a 512K empty lead-in when programming the
vendor-supplied flash.  The config file briefly mentions this.)

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Arne Georg Gleditsch]

Arne Georg Gleditsch  writes:
> Stefan Reinauer  writes:
> > The image resulting from abuild should work on real hardware.  If the
> > board comes with 1024K rom, choosing that size is a good idea.
>
> I'll do that, then.  (I am using the 512K image on real hardware today,
> but I'm padding it with a 512K empty lead-in when programming the
> vendor-supplied flash.  The config file briefly mentions this.)

Side note: the original s2912 port generates a 512K image, as far as I
can see.  Since the same board is supposed to work in both the previous
and fam10 configurations, I'm not sure how we should deal with that.

Hm, Tyan's BIOS downloads appear to be 1M as well...

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Ward Vandewege]

On Fri, Aug 22, 2008 at 02:14:30PM +0200, Arne Georg Gleditsch wrote:
> Arne Georg Gleditsch  writes:
> > Stefan Reinauer  writes:
> > > The image resulting from abuild should work on real hardware.  If the
> > > board comes with 1024K rom, choosing that size is a good idea.
> >
> > I'll do that, then.  (I am using the 512K image on real hardware today,
> > but I'm padding it with a 512K empty lead-in when programming the
> > vendor-supplied flash.  The config file briefly mentions this.)
>
> Side note: the original s2912 port generates a 512K image, as far as I
> can see.  Since the same board is supposed to work in both the previous
> and fam10 configurations, I'm not sure how we should deal with that.
>
> Hm, Tyan's BIOS downloads appear to be 1M as well...

Just change it to 1M. I had to do that for the s2881, s2882, s2891, etc. I
don't know why those were all in the tree with 512K; all the ones we have
came to us with a 1M chip.

Thanks,
Ward.

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Marc Jones]

Arne Georg Gleditsch wrote:
> Marc Jones  writes:
> > This was my fault. I missed reviewing the abuild config.
> >
> > This is now the same as serengeti_cheetah_fam10.
> >
> > A little background information. The fam10 initialization is a bit
> > larger than the K8 code. To make more room in the same 512K ROM that
> > most platforms have the normal image was dropped. The other change is
> > the use of failover. This makes the XIP CAR code live in only one
> > place instead of in the fallback and normal images.
> >
> > Arne, please test a s2912_fam10 abuild and ack this.
>
> I'll do so.  However, once I start digging into this -- would it make
> sense to fix it properly so that the target builds a full 1024K ROM (as
> the board is equipped with) and actually includes both the normal and
> failover image?

Sure, That would be great.

Marc

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Arne Georg Gleditsch]

Marc Jones  writes:
> Arne Georg Gleditsch wrote:
> > Marc Jones  writes:
> > > This was my fault. I missed reviewing the abuild config.
> > >
> > > This is now the same as serengeti_cheetah_fam10.
> > >
> > > A little background information. The fam10 initialization is a bit
> > > larger than the K8 code. To make more room in the same 512K ROM that
> > > most platforms have the normal image was dropped. The other change is
> > > the use of failover. This makes the XIP CAR code live in only one
> > > place instead of in the fallback and normal images.
> > >
> > > Arne, please test a s2912_fam10 abuild and ack this.
> >
> > I'll do so.  However, once I start digging into this -- would it make
> > sense to fix it properly so that the target builds a full 1024K ROM (as
> > the board is equipped with) and actually includes both the normal and
> > failover image?
>
> Sure, That would be great.

Sorry about the delay, I was sidetracked by other stuff.  Here's a patch
towards r3690 upping the ROM size for the S2912 Fam10 target to 1M.
Both regular and abuild images have been boot tested successfully.  (Not
having a proper abuild infrastructure set up, the abuild image fails on
not having a real payload available, but otherwise looks correct.)

Signed-off-by: Arne Georg Gleditsch

r3529 - trunk/coreboot-v2/targets/tyan/s2912_fam10

[Marc Jones]

Arne Georg Gleditsch wrote:
> Marc Jones  writes:
> > Arne Georg Gleditsch wrote:
> > > Marc Jones  writes:
> > > > This was my fault. I missed reviewing the abuild config.
> > > >
> > > > This is now the same as serengeti_cheetah_fam10.
> > > >
> > > > A little background information. The fam10 initialization is a bit
> > > > larger than the K8 code. To make more room in the same 512K ROM that
> > > > most platforms have the normal image was dropped. The other change is
> > > > the use of failover. This makes the XIP CAR code live in only one
> > > > place instead of in the fallback and normal images.
> > > >
> > > > Arne, please test a s2912_fam10 abuild and ack this.
> > >
> > > I'll do so.  However, once I start digging into this -- would it make
> > > sense to fix it properly so that the target builds a full 1024K ROM (as
> > > the board is equipped with) and actually includes both the normal and
> > > failover image?
> >
> > Sure, That would be great.
>
> Sorry about the delay, I was sidetracked by other stuff.  Here's a patch
> towards r3690 upping the ROM size for the S2912 Fam10 target to 1M.
> Both regular and abuild images have been boot tested successfully.  (Not
> having a proper abuild infrastructure set up, the abuild image fails on
> not having a real payload available, but otherwise looks correct.)
>
> Signed-off-by: Arne Georg Gleditsch

I added cross-compile to the Config-abuild.lb so it will work for me in 
with Ubuntu.

Acked-by: Marc Jones

r3711