On Wed, Nov 13, 2019 at 08:38:15AM +0100, Arnd Bergmann wrote:

On Wed, Nov 13, 2019 at 8:27 AM Christoph Hellwig [off-list ref] wrote:

quoted

On Tue, Nov 12, 2019 at 10:24:23PM +0000, Luis Chamberlain wrote:

quoted

I think this would be possible if we could flop ioremap_nocache() to UC
instead of UC- on x86. Otherwise, I can't see how we can remove this by
still not allowing direct MTRR calls.

If everything goes well ioremap_nocache will be gone as of 5.5.

As ioremap_nocache() just an alias for ioremap(), I suppose the idea would
then be to make x86 ioremap be UC instead of UC-, again matching what the
other architectures do already.

I think it's right thing to do, i.e. assume that ioremap() always does strong
UC independently on MTRR settings.

-- 
With Best Regards,
Andy Shevchenko