[PATCH v3 2/3] clk: divider: Switch from .round_rate to .determine_rate by default

[Stephen Boyd]

Quoting Martin Blumenstingl (2021-07-02 02:19:37)
> Hi Stephen,
> 
> On Fri, Jul 2, 2021 at 3:02 AM Stephen Boyd <sboyd at kernel.org> wrote:
> [...]
> > My guess is that we have drivers copying the clk_ops from the
> > divider_ops structure and so they are copying over round_rate but not
> > determine_rate.
> I just learned something new - thanks for investigating this as well!
> 
> $ git grep "clk_divider_ops\.round_rate" drivers/
> drivers/clk/bcm/clk-bcm2835.c:  return clk_divider_ops.round_rate(hw,
> rate, parent_rate);
> drivers/clk/clk-stm32f4.c:      return clk_divider_ops.round_rate(hw,
> rate, prate);
> drivers/clk/clk-stm32h7.c:      return clk_divider_ops.round_rate(hw,
> rate, prate);
> drivers/clk/clk-stm32mp1.c:             req->rate =
> clk_divider_ops.round_rate(hw, req->rate, &best_parent_rate);
> drivers/clk/imx/clk-divider-gate.c:     return
> clk_divider_ops.round_rate(hw, rate, prate);
> $ git grep "clk_divider_ro_ops\.round_rate" drivers/
> $
> 
> Changing these over to use clk_divider_ops.determine_rate doesn't seem too hard.
> The part that I am not sure about is how to organize the patches.
> 1) amend the changes to all relevant drivers (from above) to this patch
> 2) multiple patches:
> - adding .determine_rate to the default divider ops (but not removing
> .round_rate)
> - a single patch for each relevant driver (from above)
> - removing .round_rate from the default divider ops
> 
> Another approach is to first create clk_divider_determine_rate() (as
> done here) and export it.
> Then I could have one individual patch for each relevant driver (from
> above) to use:
>   .determine_rate = clk_divider_determine_rate,
> Then finally I could remove clk_divider_round_rate() and switch over
> the default divider ops to .determine_rate as well.
> 
> Which way do you prefer?
> 

I'd prefer we leave round_rate assigned in clk_divider_ops and
clk_divider_ro_ops but then also assign the determine_rate function. We
have some duplication but that's OK. Then make individual patches to
migrate each driver over to the new clk op.

We could stack a final patch on top to remove the round_rate function
from clk divider.  Unfortunately, if some driver wants to use round_rate
then it will fail in interesting ways. Probably best to live with it
until we decide to drop round_rate entirely. Patches to convert all the
round_rate code over to determine_rate would be welcome in the meantime.