[PATCH] ppc64: ppc_md.progress should not be __init

[Paul Mackerras]

Arnd Bergmann writes:

> I noticed on BPA that ppc_md.progress() is called from some places outside
> of the init call sequence, e.g. the rtas flash code. However, I copied the
> __init annotation from one of the other platform types, which caused the
> kernel to crash upon hitting the freed code segment.
> 
> I have checked that iSeries, pmac and maple all have .progress marked as
> __init and can never incorrectly hit that function after boot, so there
> is no actual breakage in the current source.
> 
> However, I find the behavior rather surprising and suggest removing this
> to have it the same way as it is required on pSeries and BPA.

What I did in the ppc32 code was to clear the ppc_md.progress pointer
when freeing the init memory.  I think that's a simpler and more
reliable fix for this problem.  If the rtas flash code wants to change
the LCD display it could call the pSeries progress routine directly.

Regards,
Paul.