[PATCH] [RFC][V3] bluegene: use MMU feature flag to conditionalize L1 writethrough code

[Eric Van Hensbergen]

On Thu, Jun 9, 2011 at 6:42 PM, Benjamin Herrenschmidt
<benh at kernel.crashing.org> wrote:
> On Thu, 2011-06-09 at 09:58 -0500, Eric Van Hensbergen wrote:
>> On Tue, Jun 7, 2011 at 7:47 PM, Benjamin Herrenschmidt
>> <benh at kernel.crashing.org> wrote:
>> > BTW. Care to explain to me why you have U2 -both- in the arguments to
>> > tlbwe and in MMUCR ? That doesn't look right to me... which one is used
>> > where and when ?
>> >
>>
>> My reading of the databook is that U2SWOAE is an enable bit that lets the U2
>> storage attribute control the behavior.
>
> You mean the MMUCR variant ?
>

Yeah, the MMCR variant acts as an enable/mask for the U2 storage attribute.

>
> Well, point is, parsing the device-tree from early boot asm is nasty,
> unless you start extending the header but that sucks. That's why I'm
> thinking it might be a good idea to look at what it takes to "convert"
> the initial entry instead, even if that involves some cache flushing
> (provided that's workable at all of course).
>

ah, okay.  I guess if its happening before the secondary cpus come up
then that should be workable.   I guess it doesn't hurt to try.

       -eric