On Tue, Jun 05, 2012 at 11:14:13AM +0100, Mark Brown wrote:
On Tue, Jun 05, 2012 at 11:08:51AM +0100, Russell King - ARM Linux wrote:
On Tue, Jun 05, 2012 at 11:03:16AM +0100, Mark Brown wrote:
I've not, but I guess there will be an issue. I'd say the fix here is on the PowerPC side - it's not good to have to have Kconfig bodges for things like this in individual drivers. Ideally we'd add clkdev support...
Or get rid of devm_clk_get()'s dependency on CLKDEV (which is in itself silly.)
That'd work too, of course (providing everyone has been implementing clk_put()). Though getting everyone on the same page with the clock API seems much more worthwhile if we can do it.
Why wouldn't clk_put() be implemented? Drivers call it just like clk_get().
The devm*() stuff should just be a layer on top of clk_get()/clk_put() and nothing more.