On Wed, Jun 28, 2017 at 05:59:49PM +0000, Priit Laes wrote:
> On Tue, Jun 27, 2017 at 11:46:07AM +0200, Maxime Ripard wrote:
> > Hi!
> > 
> > On Sun, Jun 25, 2017 at 11:45:42PM +0300, Priit Laes wrote:
> > > SATA clock on sun4i/sun7i is of type (parent) / M / 6 where
> > > 6 is fixed post-divider.
> > > 
> > > Signed-off-by: Priit Laes <plaes@plaes.org>
> > > ---
> > >  drivers/clk/sunxi-ng/ccu_div.c | 12 ++++++++++--
> > >  drivers/clk/sunxi-ng/ccu_div.h |  3 ++-
> > >  2 files changed, 12 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/drivers/clk/sunxi-ng/ccu_div.c b/drivers/clk/sunxi-ng/ccu_div.c
> > > index c0e5c10..de30e15 100644
> > > --- a/drivers/clk/sunxi-ng/ccu_div.c
> > > +++ b/drivers/clk/sunxi-ng/ccu_div.c
> > > @@ -62,8 +62,13 @@ static unsigned long ccu_div_recalc_rate(struct clk_hw *hw,
> > >  	parent_rate = ccu_mux_helper_apply_prediv(&cd->common, &cd->mux, -1,
> > >  						  parent_rate);
> > >  
> > > -	return divider_recalc_rate(hw, parent_rate, val, cd->div.table,
> > > -				   cd->div.flags);
> > > +	val = divider_recalc_rate(hw, parent_rate, val, cd->div.table,
> > > +				  cd->div.flags);
> > > +
> > > +	if (cd->common.features & CCU_FEATURE_FIXED_POSTDIV)
> > > +		val /= cd->fixed_post_div;
> > > +
> > > +	return val;
> > >  }
> > >  
> > >  static int ccu_div_determine_rate(struct clk_hw *hw,
> > > @@ -89,6 +94,9 @@ static int ccu_div_set_rate(struct clk_hw *hw, unsigned long rate,
> > >  	val = divider_get_val(rate, parent_rate, cd->div.table, cd->div.width,
> > >  			      cd->div.flags);
> > >  
> > > +	if (cd->common.features & CCU_FEATURE_FIXED_POSTDIV)
> > > +		val *= cd->fixed_post_div;
> > > +
> > >  	spin_lock_irqsave(cd->common.lock, flags);
> > >  
> > >  	reg = readl(cd->common.base + cd->common.reg);
> > 
> > You also need to adjust the round rate call back to take into account
> > the post divider before calling divider_round_rate_parent, and after
> > since that function can modify the parent_rate.
> 
> Is there a way to trigger this function? I don't see it getting called.

yep, clk_round_rate will call it, and it's also in the code path of
clk_set_rate.

> > > diff --git a/drivers/clk/sunxi-ng/ccu_div.h b/drivers/clk/sunxi-ng/ccu_div.h
> > > index 08d0744..f3a5028 100644
> > > --- a/drivers/clk/sunxi-ng/ccu_div.h
> > > +++ b/drivers/clk/sunxi-ng/ccu_div.h
> > > @@ -86,9 +86,10 @@ struct ccu_div_internal {
> > >  struct ccu_div {
> > >  	u32			enable;
> > >  
> > > -	struct ccu_div_internal		div;
> > > +	struct ccu_div_internal	div;
> > 
> > Spurious change?
> 
> Nope, it was not indented the same way as other variables.
> Should I send it as separate patch?

Hmmm, not really, leave it there.

Thanks!
Maxime

-- 
Maxime Ripard, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com