[Intel-gfx] [PATCH 04/17] drm/i915/cnl: Allow dynamic cdclk changes on CNL

Vivi, Rodrigo rodrigo.vivi at intel.com
Fri Jun 9 20:31:36 UTC 2017


On Fri, 2017-06-09 at 23:21 +0300, Imre Deak wrote:
> On Fri, Jun 09, 2017 at 11:01:17PM +0300, Vivi, Rodrigo wrote:
> > On Fri, 2017-06-09 at 22:51 +0300, Imre Deak wrote:
> > > On Fri, Jun 09, 2017 at 10:32:54PM +0300, Ville Syrjälä wrote:
> > > > On Fri, Jun 09, 2017 at 10:02:09PM +0300, Imre Deak wrote:
> > > > > On Fri, Jun 09, 2017 at 05:34:55PM +0300, Ville Syrjälä wrote:
> > > > > > On Thu, Jun 08, 2017 at 03:03:13PM -0700, Rodrigo Vivi wrote:
> > > > > > > All the low level cdclk bits are present, so let's add the required
> > > > > > > hooks to reconfigure cdclk on the fly.
> > > > > > > 
> > > > > > > v2: Rebase due to cnl_sanitize_cdclk()
> > > > > > > v3: Rebased by Rodrigo on top of Ville's cdclk rework.
> > > > > > > v4: Rebase moving cnl_calc_cdclk up to follow same order
> > > > > > >     as previous platforms.
> > > > > > > v2: Squash drm/i915/cnl: Adjust min pixel rate. to address
> > > > > > >     the current limitation where CDCLK cannot be set to 168MHz
> > > > > > >     if audio is used with 96MHz. (Imre)
> > > > > > > 
> > > > > > > Cc: Imre Deak <imre.deak at intel.com>
> > > > > > > Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> > > > > > > Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > > > > > > Signed-off-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> > > > > > > 
> > > > > > > Cannonlake also needs to adjust the minimal pixel rate
> > > > > > > as gen9 platforms. Specially for the Azalia audio case.
> > > > > > > 
> > > > > > > Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> > > > > > > Cc: Sanyog Kale <sanyog.r.kale at intel.com>
> > > > > > > Signed-off-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
> > > > > > > ---
> > > > > > >  drivers/gpu/drm/i915/intel_cdclk.c | 56 ++++++++++++++++++++++++++++++++++++--
> > > > > > >  1 file changed, 53 insertions(+), 3 deletions(-)
> > > > > > > 
> > > > > > > diff --git a/drivers/gpu/drm/i915/intel_cdclk.c b/drivers/gpu/drm/i915/intel_cdclk.c
> > > > > > > index 35a1432..1d7cfc5 100644
> > > > > > > --- a/drivers/gpu/drm/i915/intel_cdclk.c
> > > > > > > +++ b/drivers/gpu/drm/i915/intel_cdclk.c
> > > > > > > @@ -1400,6 +1400,16 @@ void bxt_uninit_cdclk(struct drm_i915_private *dev_priv)
> > > > > > >  	bxt_set_cdclk(dev_priv, &cdclk_state);
> > > > > > >  }
> > > > > > >  
> > > > > > > +static int cnl_calc_cdclk(int max_pixclk)
> > > > > > > +{
> > > > > > > +	if (max_pixclk > 336000)
> > > > > > > +		return 528000;
> > > > > > > +	else if (max_pixclk > 168000)
> > > > > > > +		return 336000;
> > > > > > > +	else
> > > > > > > +		return 168000;
> > > > > > > +}
> > > > > > > +
> > > > > > >  static void cnl_cdclk_pll_update(struct drm_i915_private *dev_priv,
> > > > > > >  				 struct intel_cdclk_state *cdclk_state)
> > > > > > >  {
> > > > > > > @@ -1641,7 +1651,7 @@ void cnl_init_cdclk(struct drm_i915_private *dev_priv)
> > > > > > >  
> > > > > > >  	cdclk_state = dev_priv->cdclk.hw;
> > > > > > >  
> > > > > > > -	cdclk_state.cdclk = 168000;
> > > > > > > +	cdclk_state.cdclk = cnl_calc_cdclk(0);
> > > > > > >  	cdclk_state.vco = cnl_cdclk_pll_vco(dev_priv, cdclk_state.cdclk);
> > > > > > >  
> > > > > > >  	cnl_set_cdclk(dev_priv, &cdclk_state);
> > > > > > > @@ -1768,7 +1778,7 @@ static int intel_max_pixel_rate(struct drm_atomic_state *state)
> > > > > > >  
> > > > > > >  		pixel_rate = crtc_state->pixel_rate;
> > > > > > >  
> > > > > > > -		if (IS_BROADWELL(dev_priv) || IS_GEN9(dev_priv))
> > > > > > > +		if (IS_BROADWELL(dev_priv) || INTEL_GEN(dev_priv) >= 9)
> > > > > > >  			pixel_rate =
> > > > > > >  				bdw_adjust_min_pipe_pixel_rate(crtc_state,
> > > > > > >  							       pixel_rate);
> > > > > > 
> > > > > > I think you might also have to adjust some of the clock limits
> > > > > > within bdw_adjust_min_pipe_pixel_rate().
> > > > > 
> > > > > Yes, we missed this. The DP/HBR2/4-lanes/audio workaround on CNL is the
> > > > > same as on GLK. So in bdw_adjust_min_pipe_pixel_rate() we could use the
> > > > > same limits as on GLK, except I don't understand the 2x multiplier there
> > > > > in
> > > > > 
> > > > > pixel_rate = max(2 * 316800, pixel_rate);
> > > > > 
> > > > > and
> > > > > 
> > > > > pixel_rate = max(2 * 2 * 96000, pixel_rate);
> > > > > 
> > > > > This pixel_rate value will be compared against CDCLK limit values in
> > > > > glk_calc_cdclk()
> > > > 
> > > > It's going to compare against 2*cdclk. But that doesn't mean it isn't
> > > > extremely confusing. To make it more sane, I think we should probably
> > > > s/min_pixclk/min_cdclk/ and actually track the "min acceptable cdclk"
> > > > rather than "the pixel rate corresponding to the min acceptable cdclk".
> > > 
> > > Arr, correct. And now I found in BSpec how to calculate cdclk for a
> > > given dotclock. So yes, the GLK part is correct, sorry for the noise.
> > > 
> > > However. For CNL the target CDCLK calculation should be the same as for
> > > GLK, based on the "GEN10 Resolution Support". So as an example there if
> > > the max pixel rate is 299MHz we can use a CDCLK of 168MHz. With the
> > > current code we'd use a CDCLK of 336MHz. Is that correct?
> > 
> > on current cnl code 2-pixel-per-clock is disabled. I got a conflict and
> > decided to take a look after upstreaming the basic support.
> > 
> > So I believe the current code is the correct one, right?! (But honestly
> > I'm a bit lost here)
> 
> Ok, found that internal patch now, with that the current code looks ok.
> Would make sense to add a note about this to the commit message.

What about: "Note: Current CNL code doesn't support 2 pixel per clock.
All 2*cdclk that is currently in place for GLK doesn't apply for CNL for
now."

with that can I keep your rv-b? and apply the comment when merging?

> 
> > > 
> > > > 
> > > > > not CD2XCLK limit values. Removing the 2x multiplier
> > > > > from the above pixel_rate values wouldn't change anything on GLK AFAICS
> > > > > and it would allow us to use the same code to calculate these on GLK
> > > > > and CNL (and simplify the second pixel_rate calculation).
> > 
> > > > > 
> > > > > --Imre
> > > > > 
> > > > > > 
> > > > > > > @@ -1929,6 +1939,40 @@ static int bxt_modeset_calc_cdclk(struct drm_atomic_state *state)
> > > > > > >  	return 0;
> > > > > > >  }
> > > > > > >  
> > > > > > > +static int cnl_modeset_calc_cdclk(struct drm_atomic_state *state)
> > > > > > > +{
> > > > > > > +	struct drm_i915_private *dev_priv = to_i915(state->dev);
> > > > > > > +	struct intel_atomic_state *intel_state =
> > > > > > > +		to_intel_atomic_state(state);
> > > > > > > +	int max_pixclk = intel_max_pixel_rate(state);
> > > > > > > +	int cdclk, vco;
> > > > > > > +
> > > > > > > +	cdclk = cnl_calc_cdclk(max_pixclk);
> > > > > > > +	vco = cnl_cdclk_pll_vco(dev_priv, cdclk);
> > > > > > > +
> > > > > > > +	if (cdclk > dev_priv->max_cdclk_freq) {
> > > > > > > +		DRM_DEBUG_KMS("requested cdclk (%d kHz) exceeds max (%d kHz)\n",
> > > > > > > +			      cdclk, dev_priv->max_cdclk_freq);
> > > > > > > +		return -EINVAL;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	intel_state->cdclk.logical.vco = vco;
> > > > > > > +	intel_state->cdclk.logical.cdclk = cdclk;
> > > > > > > +
> > > > > > > +	if (!intel_state->active_crtcs) {
> > > > > > > +		cdclk = cnl_calc_cdclk(0);
> > > > > > > +		vco = cnl_cdclk_pll_vco(dev_priv, cdclk);
> > > > > > > +
> > > > > > > +		intel_state->cdclk.actual.vco = vco;
> > > > > > > +		intel_state->cdclk.actual.cdclk = cdclk;
> > > > > > > +	} else {
> > > > > > > +		intel_state->cdclk.actual =
> > > > > > > +			intel_state->cdclk.logical;
> > > > > > > +	}
> > > > > > > +
> > > > > > > +	return 0;
> > > > > > > +}
> > > > > > > +
> > > > > > >  static int intel_compute_max_dotclk(struct drm_i915_private *dev_priv)
> > > > > > >  {
> > > > > > >  	int max_cdclk_freq = dev_priv->max_cdclk_freq;
> > > > > > > @@ -1960,7 +2004,9 @@ static int intel_compute_max_dotclk(struct drm_i915_private *dev_priv)
> > > > > > >   */
> > > > > > >  void intel_update_max_cdclk(struct drm_i915_private *dev_priv)
> > > > > > >  {
> > > > > > > -	if (IS_GEN9_BC(dev_priv)) {
> > > > > > > +	if (IS_CANNONLAKE(dev_priv)) {
> > > > > > > +		dev_priv->max_cdclk_freq = 528000;
> > > > > > > +	} else if (IS_GEN9_BC(dev_priv)) {
> > > > > > >  		u32 limit = I915_READ(SKL_DFSM) & SKL_DFSM_CDCLK_LIMIT_MASK;
> > > > > > >  		int max_cdclk, vco;
> > > > > > >  
> > > > > > > @@ -2157,6 +2203,10 @@ void intel_init_cdclk_hooks(struct drm_i915_private *dev_priv)
> > > > > > >  		dev_priv->display.set_cdclk = skl_set_cdclk;
> > > > > > >  		dev_priv->display.modeset_calc_cdclk =
> > > > > > >  			skl_modeset_calc_cdclk;
> > > > > > > +	} else if (IS_CANNONLAKE(dev_priv)) {
> > > > > > > +		dev_priv->display.set_cdclk = cnl_set_cdclk;
> > > > > > > +		dev_priv->display.modeset_calc_cdclk =
> > > > > > > +			cnl_modeset_calc_cdclk;
> > > > > > >  	}
> > > > > > >  
> > > > > > >  	if (IS_CANNONLAKE(dev_priv))
> > > > > > > -- 
> > > > > > > 1.9.1
> > > > > > 
> > > > > > -- 
> > > > > > Ville Syrjälä
> > > > > > Intel OTC
> > > > 
> > > > -- 
> > > > Ville Syrjälä
> > > > Intel OTC
> > 



More information about the Intel-gfx mailing list