Re: [Intel-gfx] [PATCH 2/3] drm/i915: Do vblank evasion correctly if vrr push has already been sent

["Navare, Manasi" <manasi.d.navare@intel.com>]

On Wed, Nov 17, 2021 at 03:09:45PM -0800, Navare, Manasi wrote:
> On Wed, Nov 17, 2021 at 11:04:05PM +0200, Ville Syrjälä wrote:
> > On Wed, Nov 17, 2021 at 01:10:13PM -0800, Navare, Manasi wrote:
> > > On Wed, Nov 17, 2021 at 08:31:02PM +0200, Ville Syrjala wrote:
> > > > From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > > > 
> > > > Let's adjust the vblank evasion to account for the case where
> > > > a push has already been sent. In that case the vblank exit will start
> > > > at vmin vblank start (as opposed to vmax vblank start when no push
> > > > has been sent).
> > > > 
> > > > This should minimize the effects of the tiny race between sampling
> > > > the frame counter vs. intel_vrr_send_push() during the previous frame.
> > > > This will also be required if we want to do mailbox style updates with
> > > > vrr since then we'd definitely do multiple commits per frame. Currently
> > > > mailbox updates are only used by the legacy cursor, but we don't do
> > > > vrr push for those.
> > > > 
> > > > Cc: Manasi Navare <manasi.d.navare@intel.com>
> > > > Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

I still want to see if PUSH send can be cleared if its coming after Vmax because then that
will affect the next frame that will always just terminate at Vmin. But we can continue that
discussion later. For this patch

Reviewed-by: Manasi Navare <manasi.d.navare@intel.com>

Manasi

> > > > ---
> > > >  drivers/gpu/drm/i915/display/intel_crtc.c | 10 +++++++---
> > > >  drivers/gpu/drm/i915/display/intel_vrr.c  | 12 ++++++++++++
> > > >  drivers/gpu/drm/i915/display/intel_vrr.h  |  1 +
> > > >  3 files changed, 20 insertions(+), 3 deletions(-)
> > > > 
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_crtc.c b/drivers/gpu/drm/i915/display/intel_crtc.c
> > > > index cf403be7736c..eb5444f90e77 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_crtc.c
> > > > +++ b/drivers/gpu/drm/i915/display/intel_crtc.c
> > > > @@ -470,10 +470,14 @@ void intel_pipe_update_start(struct intel_crtc_state *new_crtc_state)
> > > >  	if (intel_crtc_needs_vblank_work(new_crtc_state))
> > > >  		intel_crtc_vblank_work_init(new_crtc_state);
> > > >  
> > > > -	if (new_crtc_state->vrr.enable)
> > > > -		vblank_start = intel_vrr_vmax_vblank_start(new_crtc_state);
> > > > -	else
> > > > +	if (new_crtc_state->vrr.enable) {
> > > > +		if (intel_vrr_is_push_sent(new_crtc_state))
> > > > +			vblank_start = intel_vrr_vmin_vblank_start(new_crtc_state);
> > > 
> > > Actually if Push is sent then the vblank gets terminated at that point which falls between vmin and vmax so
> > > not necessarily at Vmin but at anytime between vmin and vmax. Is it correct to calculate it based on vmin?
> > 
> > If you do a push between vmin and vmax then the vblank terminates
> > immediately and the PUSH_SEND bit gets cleared. The only way the bit
> > stays set is you set it after vmax/before vmin.
> >
> 
> The scenario where PUSH comes in during the active so that its before Vmin, then it will stay on until we hit vmin to
> terminate the vblank.
> But when we reach Vmax, the HW will terminate the vblank anyways, should we actually alter the code to not
> allow any PUSH after Vmax?
> 
> Manasi
>  
> > > 
> > > > +		else
> > > > +			vblank_start = intel_vrr_vmax_vblank_start(new_crtc_state);
> > > > +	} else {
> > > >  		vblank_start = intel_mode_vblank_start(adjusted_mode);
> > > > +	}
> > > >  
> > > >  	/* FIXME needs to be calibrated sensibly */
> > > >  	min = vblank_start - intel_usecs_to_scanlines(adjusted_mode,
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_vrr.c b/drivers/gpu/drm/i915/display/intel_vrr.c
> > > > index c335b1dbafcf..db1c3902fc2d 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_vrr.c
> > > > +++ b/drivers/gpu/drm/i915/display/intel_vrr.c
> > > > @@ -193,6 +193,18 @@ void intel_vrr_send_push(const struct intel_crtc_state *crtc_state)
> > > >  		       TRANS_PUSH_EN | TRANS_PUSH_SEND);
> > > >  }
> > > >  
> > > > +bool intel_vrr_is_push_sent(const struct intel_crtc_state *crtc_state)
> > > > +{
> > > > +	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
> > > > +	struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
> > > > +	enum transcoder cpu_transcoder = crtc_state->cpu_transcoder;
> > > > +
> > > > +	if (!crtc_state->vrr.enable)
> > > > +		return false;
> > > > +
> > > > +	return intel_de_read(dev_priv, TRANS_PUSH(cpu_transcoder)) & TRANS_PUSH_SEND;
> > > 
> > > But HW clears this bit after double buffer update. Is this a good inidcation to check the PUSH_SEND bit?
> > > 
> > > Manasi
> > > 
> > > > +}
> > > > +
> > > >  void intel_vrr_disable(const struct intel_crtc_state *old_crtc_state)
> > > >  {
> > > >  	struct intel_crtc *crtc = to_intel_crtc(old_crtc_state->uapi.crtc);
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_vrr.h b/drivers/gpu/drm/i915/display/intel_vrr.h
> > > > index 96f9c9c27ab9..1c2da572693d 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_vrr.h
> > > > +++ b/drivers/gpu/drm/i915/display/intel_vrr.h
> > > > @@ -23,6 +23,7 @@ void intel_vrr_compute_config(struct intel_crtc_state *crtc_state,
> > > >  void intel_vrr_enable(struct intel_encoder *encoder,
> > > >  		      const struct intel_crtc_state *crtc_state);
> > > >  void intel_vrr_send_push(const struct intel_crtc_state *crtc_state);
> > > > +bool intel_vrr_is_push_sent(const struct intel_crtc_state *crtc_state);
> > > >  void intel_vrr_disable(const struct intel_crtc_state *old_crtc_state);
> > > >  void intel_vrr_get_config(struct intel_crtc *crtc,
> > > >  			  struct intel_crtc_state *crtc_state);
> > > > -- 
> > > > 2.32.0
> > > > 
> > 
> > -- 
> > Ville Syrjälä
> > Intel