[PATCH v2 0/9] ARM: multi-platform kconfig cleanup and mach-virt removal

[Arnd Bergmann]

On Wednesday 12 February 2014 08:07:12 Rob Herring wrote:
> On Wed, Feb 12, 2014 at 7:46 AM, Will Deacon <will.deacon at arm.com> wrote:
> > On Wed, Feb 12, 2014 at 01:26:41PM +0000, Arnd Bergmann wrote:
> >> On Tuesday 11 February 2014, Rob Herring wrote:
> >> > The previous version [1] was mainly a discussion about v6 vs. v6K.
> >> > Several platforms have this wrong and incorrectly select v6 when the
> >> > more optimal v6K option could be used. After more research, my memory
> >> > about i.MX31 was wrong and it does need to remain v6.
> >>
> >> Just curious: do you have more information on this? Are all i.MX31 ARMv6
> >> and all i.MX35 v6k as the current Kconfig claims,  or is it more
> >> complicated?
> >
> > Slightly tangential, but the one to watch out for is 1136. Prior to r1 (i.e.
> > r0pX), it is v6 but r1pX+ are v6k (without SMP).
> 
> Right. I originally was thinking that the MX31 1.x was r0pX and MX31
> 2.x was r1pX and that there are no 1.x chips around. However, after
> checking the errata sheet, 1.x is r0p1 and 2.x is r0p4. It must have
> been one of the other 1136 chips we did that moved to r1pX.

Ok.

> >> * integrator and realview apparently allow both CPU_V6 and CPU_V6K
> >>   to be manually selected. Is that actually the correct behavior
> >>   in that both kinds of core tiles exist?
> >
> > I have 1136 r0p1 on an integrator CP, so I suppose it could also take an
> > 1136 r1pX without any trouble.
> 
> Presumably some of both exist which is why the config options are as they are?

Possible, but I wouldn't trust that part of the kernel very far. Especially
the MMU-less CPUs in there seem to be listed a bit randomly.

On the topic of V6, we don't support CPU_V6T2 at all, though I assume that
there is an ARM1156 core tile for integrator and realview, and we support
running the V6/V6K parts with MMU turned off. Would all revisions of ARM1156
work with CPU_V6K and !MMU?

	Arnd