[Cbe-oss-dev] [RFC,PATCH 1/1] Allow PCAB to boot Linux 2.6.19

[Jean-Christophe Dubois]

On Thursday 16 November 2006 01:14, Benjamin Herrenschmidt wrote:
> On Wed, 2006-11-15 at 18:17 +0100, Jean-Christophe Dubois wrote:
> > SLOF on FACB/PCAB/CAB still has a quite mixed up device tree. The latest
> > patch for 2.6.19 will not allow linux to boot on the PCAB as it will try
> > to access the Axon MPIC with the ioremap method (instead of the DCR
> > method) but it is then missing the 16 bytes striding info to go with it.
>
> The device-tree needs fixing then :-)

I agree but so far I did not get a lot of commitment on this. At least not 
short term. I am stuck at an unofficial 2.6.16 ...

> > This patch allow the PCAB to boot 2.6.19 with Arnd patches from November
> > 11th.
>
> Well, I implemented the native DCR support in MPIC to avoid having to do
> that ...

I understand this but unfortunately it doesn't handle the "old way" in the 
xCAB tree and make the board "Machine check" at first MPIC access.

> > Note that unfortunately the MPIC controller info is present twice in the
> > actual PCAB device tree. Therefore the MPIC is initialized twice. Once
> > with the ioremap method and once with the DCR method. It is obviously not
> > really the best thing.
> >
> > A cleaner method would be to update SLOF on FCAB/PCAB/CAB so that it is
> > at the same level as the Malta blade but we are not here yet.
>
> Another is to add a hack to "skip" the wrong device-node when probing
> for MPICs for now and make sure it only uses the DCR node. (Note that
> the last time I tested on SIMICS, I had a SLOF image that had a proper
> MPIC node...)

I believe that the FCAB doesn't have the "good node" (I'll check when I get to 
the office) so it is a bit more cumbersome thant this depending on the board 
type. Also for the FCAB I am not sure what kind of commitment I could get to 
have its firmware ever upgraded.

> > Also this patch doesn't allow yet the ethernet driver to work on xCAB for
> > 2.6.19.
> >
> > Signed-off-by: Jean-Christophe Dubois <jdubois at mc.com>
>
> I don't see any good reason to merge that patch though. SLOF will be
> fixed soon enough and there are better workarounds anyway since you say
> the tree contains the proper DCR-based MPIC node. Just skip the wrong
> one for now from the platform code. I don't think that patch nor the
> workaround needs to be merged in any tree anyway as the firmware will be
> fixed.

Unless I get some commitment to update FCAB and PCAB firmware to the correct 
level, this patch is needed for me. 

> Regarding EMAC, there seem to be other issues that Murali reported to
> me. I need to do a new patch with the workarounds he sent me along with
> other cleanups. Hopefully, next week.

OK

>
> Cheers,
> Ben.