On Wed, Aug 17, 2011 at 09:51:31AM +0200, Linus Walleij wrote:

I would suggest that in all patches using these functions, try
to replace:

mmc_host_clk_disable() -> mmc_host_clk_ungate()
mmc_host_clk_enable() -> mmc_host_clk_gate()

Wow, that is indeed *much* cleaner way of doing this!

One thing is that if I call these from those ios functions,
mmc_host_clk_ungate() will always try to restore the clock
even if there is really no need. Do you see this as a problem?

Please tell us if this works!

Certainly. I'll try this overnight and see whether it works.

I understand that the names can be a bit confusing by but
I think you can convince yourself that what this will do is
simply increase the refcount host->clk_requests so the
clock is not gated across these sections.

If you think the names of the functions are confusing then
you may rename them, say like this:

mmc_host_clk_ungate() -> mmc_host_clk_hold()
mmc_host_clk_gate() -> mmc_host_clk_release()

Which would make the usecases more clear, I'd be happy
to ACK a patch for this.

I agree, I'll cook a patch for that also.

Thanks for the comments.