[PATCH v2 6/7] Holly DTS

Josh Boyer jwboyer at linux.vnet.ibm.com
Sun May 6 07:36:12 EST 2007


On Sat, 2007-05-05 at 22:44 +0200, Segher Boessenkool wrote:
> >>> +/ {
> >>> +	model = "ppc750-tsi109";
> >>
> >> "model" should be something really specific; typically
> >> a (unique!) model number.  This means you can't use the
> >> same device tree for Holly and Hickory (but there are
> >> more reasons for that; see below).
> >>
> >>> +	compatible = "ppc750-tsi";
> >>
> >> The needs to be more specific as well; "ibm,holly"
> >> or something.
> >
> > Why?  Holly and Hickory share the same memory map and devices.  The 
> > only
> > thing that differs from what I can tell is the actual CPU itself.
> 
> And the model number I suspect.
> 
> "ppc750-tsi" really isn't good enough as a "compatible"
> property, there are many many more boards with some 750
> and some TSI bridge.

Ok.  It really doesn't matter one way or another to me, since it's
something I'm defining anyway.  "ibm,holly" and/or "ibm,hickory" or
"ibm,ppc750-tsi" would likely work.

Actually, I might muck around with the PVR in the bootwrapper and poke
the correct model and compatible nodes in there.  That might work out
well enough.

> >>> +  	tsi109 at c0000000 {
> >>> +		device_type = "tsi-bridge";
> >>
> >> Don't put a "device_type" here, it is useless
> >> (and undefined).  There are more like this, but
> >> perhaps Linux (wrongly) probes on "device_type"
> >> for those, so the kernel would need updating
> >> first.
> >
> > It's not useless.  The TSI code probes by device_type all over the
> > place.  Do you have an example of how to probe for something like this
> > without using device_type?
> 
> Sure: use "compatible" instead.
> 
> This problem is all over the place, don't worry
> about it too much.  It would be good if new ports
> could stop adding to the madness though ;-)

For now, I'm inclined to leave it as is to minimize the amount of things
needed to actually get Holly merged.  However, I'll start looking at
doing it via compatible shortly.  Hopefully someone will be able to test
out the mpc7448hpc2 board as I go to make sure things are breaking
there.

> >>> +	  	MPIC: pic at 7400 {
> >>> +			built-in;
> >>
> >> Why say this for MPIC only, not for most other nodes?
> >> What binding defines "built-in", anyway?
> >
> > http://playground.sun.com/1275/bindings/chrp/chrp1_8a.ps
> 
> But this isn't a CHRP compatible board, that binding
> doesn't apply.

I'm not pretending it is.  You just asked which binding defined it and I
was proud of myself for actually finding one ;).

It doesn't actually do anything in this case, so it can be removed.

josh




More information about the Linuxppc-dev mailing list