[BUG] Revert 0b05e2d671c4 'powerpc/32: cacheable_memcpy becomes memcpy'

[Scott Wood]

On Mon, 2015-09-21 at 17:10 +1000, Michael Ellerman wrote:
> On Fri, 2015-09-18 at 16:45 +0200, Thomas Gleixner wrote:
> > On Fri, 18 Sep 2015, Segher Boessenkool wrote:
> > 
> > > On Fri, Sep 18, 2015 at 08:01:57PM +1000, Michael Ellerman wrote:
> > > > Weird.
> > > > 
> > > > Can you try:
> > > > 
> > > >   d690740f22f6 ("powerpc/powernv: Enable LEDS support")
> > > > 
> > > > That is the commit before my merge of Scott's tree, which had all the 
> > > > Freescale
> > > > & 32-bit stuff.
> > > > 
> > > > If that works, it would isolate it to only about 30 commits from 
> > > > Scott's tree.
> > > 
> > > My theory is it is not a kernel problem at all; instead, uboot 
> > > overwrites
> > > some little part of the kernel image.
> > 
> > u-boot is just loading a cuImage which is a kernel with appended
> > device tree. So u-boot is out of the picture,
> 
> Unless something is still DMA'ing after u-boot has handed over to the 
> kernel.
> We had that happen on OF years ago.
>
> > still investigating (when time allows)
> 
> Thanks. I assume no one at Freescale has been able to reproduce this?

I don't have an mpc52xx to test on.  I can boot a p4080 with tracing enabled.

-Scott