Re: [PATCH v3 0/7] mv643xx.c: Add basic device tree support.

On Mon, 2012-09-10 at 14:22 +0000, Arnd Bergmann wrote:

Following up on the old discussion, I talked briefly about this
issue with BenH at the kernel summit. The outcome basically is that
it's a bit sad to have incompatible bindings, but it's not the end
of the world,and it's more important to do it right this time.

Just make sure that you use different values for the 'compatible'
strings and then do what you need to get the ARM hardware working.

Ideally, the new binding should be written in a way that powerpc
machines can use the same one, but the existing ones all use
an version of Open Firmware that is not going to get updated
and it's also not too likely that we are going to see new
powerpc machines based on this chip.

Right, mostly these machines where the Pegasos. Those came with a fairly
busted variant of Open Firmware which generated a pretty gross
device-tree.

For some reason, the manufacturer of those things was never willing to
fix anything in their firmware (despite the distributor providing
patches etc...), seemingly on the assumption that whatever they were
doing was perfect and operating system people like us didn't matter one
little bit :-)

So I don't care much about it. It would be nice to keep them working
since people in the community still have them but if it goes through
some "compat" code that detects old/broken device-trees and eventually
disappears when we finally drop support, then so be it.

Cheers,
Ben.

Pong.  ;-)

On Tue, Sep 11, 2012 at 04:03:31PM +1000, Benjamin Herrenschmidt wrote:

On Mon, 2012-09-10 at 14:22 +0000, Arnd Bergmann wrote:

quoted

Following up on the old discussion, I talked briefly about this
issue with BenH at the kernel summit. The outcome basically is that
it's a bit sad to have incompatible bindings, but it's not the end
of the world,and it's more important to do it right this time.

Just make sure that you use different values for the 'compatible'
strings and then do what you need to get the ARM hardware working.

Ideally, the new binding should be written in a way that powerpc
machines can use the same one, but the existing ones all use
an version of Open Firmware that is not going to get updated
and it's also not too likely that we are going to see new
powerpc machines based on this chip.

Right, mostly these machines where the Pegasos. Those came with a fairly
busted variant of Open Firmware which generated a pretty gross
device-tree.

For some reason, the manufacturer of those things was never willing to
fix anything in their firmware (despite the distributor providing
patches etc...), seemingly on the assumption that whatever they were
doing was perfect and operating system people like us didn't matter one
little bit :-)

So I don't care much about it. It would be nice to keep them working
since people in the community still have them but if it goes through
some "compat" code that detects old/broken device-trees and eventually
disappears when we finally drop support, then so be it.

Ian,

What is the status of this work?

thx,

Jason.