Re: [PATCH 03/18] drm/tidss: Add mode_fixup to adjust the clock based on HW

[Tomi Valkeinen]

Hi,

On 21/03/2025 16:17, Maxime Ripard wrote:
> On Thu, Mar 20, 2025 at 05:59:58PM +0200, Tomi Valkeinen wrote:
> > At the moment the driver just sets the clock rate with clk_set_rate(),
> > and if the resulting rate is not the same as requested, prints a debug
> > print, but nothing else.
> >
> > Add mode_fixup(), in which the clk_round_rate() is used to get the
> > "rounded" rate, and set that to the adjusted_mode.
> >
> > In practice, with the current K3 SoCs, the display PLL is capable of
> > producing very exact clocks, so most likely the rounded rate is the same
> > as the original one.
> >
> > Signed-off-by: Tomi Valkeinen <tomi.valkeinen@xxxxxxxxxxxxxxxx>
> > ---
> >   drivers/gpu/drm/tidss/tidss_crtc.c  | 22 ++++++++++++++++++++++
> >   drivers/gpu/drm/tidss/tidss_dispc.c |  6 ++++++
> >   drivers/gpu/drm/tidss/tidss_dispc.h |  2 ++
> >   3 files changed, 30 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/tidss/tidss_crtc.c b/drivers/gpu/drm/tidss/tidss_crtc.c
> > index 1604eca265ef..b3338dac25bc 100644
> > --- a/drivers/gpu/drm/tidss/tidss_crtc.c
> > +++ b/drivers/gpu/drm/tidss/tidss_crtc.c
> > @@ -309,7 +309,29 @@ enum drm_mode_status tidss_crtc_mode_valid(struct drm_crtc *crtc,
> >   	return dispc_vp_mode_valid(tidss->dispc, tcrtc->hw_videoport, mode);
> >   }
> >   
> > +static bool tidss_crtc_mode_fixup(struct drm_crtc *crtc,
> > +				  const struct drm_display_mode *mode,
> > +				  struct drm_display_mode *adjusted_mode)
> > +{
> > +	struct tidss_crtc *tcrtc = to_tidss_crtc(crtc);
> > +	struct drm_device *ddev = crtc->dev;
> > +	struct tidss_device *tidss = to_tidss(ddev);
> > +	long rate;
> > +
> > +	rate = dispc_vp_round_clk_rate(tidss->dispc, tcrtc->hw_videoport,
> > +				       adjusted_mode->clock * 1000);
> > +	if (rate < 0)
> > +		return false;
> > +
> > +	adjusted_mode->clock = rate / 1000;
> > +
> > +	drm_mode_set_crtcinfo(adjusted_mode, 0);
> > +
> > +	return true;
> > +}
>
> mode_fixup is deprecated in favor of atomic_check. If you can't use it
> for some reason, it should be documented one way or another.

Ok. I didn't see a note about it being deprecated (just optional), so I 
went with fixup. Atomic check is called for every commit, and this 
should be only done before enabling a ctrc. I need to dig that 
"going-to-be-enabled" from the state.

 Tomi