Re: Of clk_hw_unregister_fractional_divider

[Stephen Boyd <sboyd@kernel.org>]

Quoting Dr. David Alan Gilbert (2024-10-09 17:58:21)
> * Stephen Boyd (sboyd@kernel.org) wrote:
> > Quoting Dr. David Alan Gilbert (2024-10-08 17:54:24)
> > > Hi,
> > >   One of my scripts noticed that clk_hw_unregister_fractional_divider
> > > was unused and I was thinking of dead coding it, but thought it worth
> > > asking whether it's actually a missing call to it.
> > > 
> > >   It was added in 2016 by:
> > > 
> > > commit 39b44cff4ad4af6d7abd9dd2acb288b005c26503
> > > Author: Stephen Boyd <sboyd@codeaurora.org>
> > > Date:   Sun Feb 7 00:15:09 2016 -0800
> > > 
> > >     clk: fractional-divider: Add hw based registration APIs
> > >     
> > > and I can see clk_hw_register_fractional_divider is called from
> > > clk_register_fractional_divider.
> > > 
> > > Is that just consequence of them being system devices that never
> > > get unregistered or something else?
> > > 
> > 
> > I don't know what you mean by system devices.
> 
> I mean something that can't get unplugged/removed.

Ok.

> 
> > The clk_hw prefixed functions and clk_hw based APIs should be used
> > instead of struct clk APIs for clk providers in general. I'd accept a
> > patch that migrates the two calls to clk_register_fractional_divider()
> > with a call to clk_hw_register_fractional_divider() instead. If that's
> > done we can remove clk_register_fractional_divider(), which is the
> > overall plan but nobody got around to it so far.
> 
> Oh hmm, that doesn't look too hard for those two - although I'm not sure
> I have a way of testing it?

Sure. I don't think anything really changes if that's done so probably a
compile test is all that's needed.

> 
> However, my question was about the clk_hw_*un*register_fractional_divider
> - I can see the clk_register_fractional_divider can be replaced by
> clk_hw_register_fractional_divider - but would there need to be
> a call to the unregister somewhere? (Which is what my script noticed
> there's no callers of).
> 

Yes they would call clk_hw_unregister_fractional_divider() as well after
changing to use clk_hw pointers for the registration path.