Re: [PATCH v3 1/7] clk: divider: add flag to limit possible dividers to even numbers

["Heiko Stübner" <heiko@sntech.de>]

Am Dienstag, 11. Juni 2013, 20:57:50 schrieb Mike Turquette:
> Quoting Heiko Stübner (2013-06-11 04:29:32)
> 
> > SoCs like the Rockchip Cortex-A9 ones contain divider some clocks
> > that use the regular mechanisms for storage but allow only even
> > dividers and 1 to be used.
> > 
> > Therefore add a flag that lets _is_valid_div limit the valid dividers
> > to these values. _get_maxdiv is also adapted to return even values
> > for the CLK_DIVIDER_ONE_BASED case.
> > 
> > Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> > ---
> > 
> >  drivers/clk/clk-divider.c    |   14 ++++++++++++--
> >  include/linux/clk-provider.h |    2 ++
> >  2 files changed, 14 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/clk/clk-divider.c b/drivers/clk/clk-divider.c
> > index ce5cfe9..bdee7cf 100644
> > --- a/drivers/clk/clk-divider.c
> > +++ b/drivers/clk/clk-divider.c
> > @@ -45,8 +45,16 @@ static unsigned int _get_table_maxdiv(const struct
> > clk_div_table *table)
> > 
> >  static unsigned int _get_maxdiv(struct clk_divider *divider)
> >  {
> > 
> > -       if (divider->flags & CLK_DIVIDER_ONE_BASED)
> > -               return div_mask(divider);
> > +       if (divider->flags & CLK_DIVIDER_ONE_BASED) {
> > +               unsigned int div = div_mask(divider);
> > +
> > +               /* decrease to even number */
> > +               if (divider->flags & CLK_DIVIDER_EVEN)
> > +                       div--;
> > +
> > +               return div;
> > +       }
> > +
> > 
> >         if (divider->flags & CLK_DIVIDER_POWER_OF_TWO)
> >         
> >                 return 1 << div_mask(divider);
> >         
> >         if (divider->table)
> > 
> > @@ -141,6 +149,8 @@ static bool _is_valid_div(struct clk_divider
> > *divider, unsigned int div)
> > 
> >                 return is_power_of_2(div);
> >         
> >         if (divider->table)
> >         
> >                 return _is_valid_table_div(divider->table, div);
> > 
> > +       if (divider->flags & CLK_DIVIDER_EVEN && div != 1 && (div % 2) !=
> > 0)
> 
> Is it correct to check for 'div != 1' here?  Wouldn't that check only be
> valid in the presence of CLK_DIVIDER_ONE_BASED?
> 
> Maybe something like this would be more correct:
> 
> 	if (divider->flags & CLK_DIVIDER_EVEN && (div % 2) != 0) {
> 		if (divider->flags & CLK_DIVIDER_ONE_BASED && div == 1)
> 			return true;
> 		return false;
> 	}

hmm, not necessarily. As I understand the doc DIVIDER_ONE_BASED describes how 
the divider is stored in the register, i.e. if the register value is div-1 or 
div.

When div is 1, it of course means don't divide [rate/1], and CLK_DIVIDER_EVEN 
needs to just make sure that bigger dividers that really divide the rate are 
even values (so limiting the possible dividers to 1 2 4 ...), independent on 
how the divider value is stored in the register.