On Thu, Jul 28, 2016 at 07:55:10PM +0100, Mark Brown wrote:

quoted

quoted

This device seems to be used on x86 systems so we'll need to ensure that
they register clocks for this.  They really should set this up using
quirks keyed off DMI information or similar so it's hidden from other
systems.

quoted

Hmm..the change defines this mclk as a optional property so I'm
not sure how it would affect x86 systems. (Would love to refine

Ah, clk_prepare_enable() silently ignores NULL as an argument?  It is a
bit messy to do things that way but might be the most practical thing.

At least the one in the drivers/clk/clk.c does:

660 int clk_prepare(struct clk *clk)
661 {
662         int ret;
663
664         if (!clk)
665                 return 0;
...
774 int clk_enable(struct clk *clk)
775 {
776         unsigned long flags;
777         int ret;
778
779         if (!clk)
780                 return 0;

Hopefully they'll speak up - we're in the merge window anyway so it'll
be just over a week before this can go into -next.

I see. Let's wait then.

Thanks
Nicolin