On Tue, May 13, 2014 at 08:01:05AM +1000, Benjamin Herrenschmidt wrote: > On Mon, 2014-05-12 at 14:27 -0500, Rob Herring wrote: > > But what you wrote was for unflattened tree. For early FDT uses, do we > > really need to worry about PCI or other special cases? The current FDT > > address translation code in arch/powerpc/boot/devtree.c (yes, now we > > have 3 implementations) does not for example. > > This is specific to a bootloader case yes and it's ... fishy. > > PCI on ARM is here now. PCI is everywhere (well, pseudo-PCI but that's > the basic issue here) on Intel embedded. > > Some embedded systems will probably want to enumerate from the firmware > and Linux will need to translate. For example because the UART or some > critical GPIO that needs to be handled early on is behind a pseudo-PCI > interface of some sort. > > I think we need to handle these cases rather than fail in obscure ways.
Yeah, I think PCI is way to common to just ignore.
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
pgpEoB7FvbJJt.pgp
Description: PGP signature
