[PATCH 09/15] [POWERPC] 86xx: Add uli1575 pci-bridge sector to MPC8641HPCN dts file.

[Segher Boessenkool]

>>> This can't be right, and I suspect it will break any kernel access to
>>> the first 0x1000 bytes of the CCSR_BAR space.  reg should actually
>>> describe the register space of the SOC.  If ranges needs to specify
>>> that, too, they should be able to be redundant.  But this looks like
>>> a big hack, to me.  Am I missing something?
>>
>> "reg" and "ranges" can never overlap really.
>
> Is this really true, always?

Well think about it.  "reg" is the addresses your bridge
device's registers sit at.  "ranges" is the addresses your
bridge translates to its child bus.

> What about something like a PReP PCI bridge, which has the registers
> for indirect config-space access in the same range as the ISA IO
> space?

Yeah there are always nasty corner cases that are hard
to express.  Like a PHB that is visible as a device on
its own PCI bus -- it would have to be its own child in
the device tree!  Such devices are best dealt with on
a one-by-one basis.


Segher