Avi Kivity wrote:
> Jes Sorensen wrote:
>> Hi,
>>
>> I am not crazy about this patch. You need to use cpy_physical_memory_rw()
>> in the hob and nvram code too, not just in the ipf.c code.
>>
>> What about the flush_icache_range() call you removed - is it safe to
>> just discard that?
>>
>> I was in the process of working through this myself, but I am not
>> quite finished. If you don't mind waiting a couple hours, I should
>> have something a fair bit simpler to solve the same problem.
>>
>> Biggest issue is the flush_icache_range() one.
>>
> I haven't pushed this out yet, so I can apply a replacement patch.

Hi,

As promised here is my proposal for how to implement this patch. I have
made the hob and nvram code use cpu_physical_memory_{read,write} as
well, handled the cache flushing address pointer issue in the flush
function, which needs some improvement to be done later.

By doing this I have been able to remove a bunch of globals and a number
of variables being passed around which we really don't need.

Xiantao, what do you think of this solution?

Cheers,
Jes