Re: [Intel-gfx] [PATCH v4 4/6] drm/i915/dp: Fix MST disable sequences

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Wed, Dec 18, 2019 at 09:25:33PM +0000, Souza, Jose wrote:
> On Wed, 2019-12-18 at 22:08 +0200, Ville Syrjälä wrote:
> > On Wed, Dec 18, 2019 at 10:59:08AM -0800, José Roberto de Souza
> > wrote:
> > > The disable sequence after wait for transcoder off was not
> > > correctly
> > > implemented.
> > > The MST disable sequence is basically the same for HSW, SKL, ICL
> > > and
> > > TGL, with just minor changes for TGL.
> > > 
> > > With this last patch we finally fixed the hotplugs triggered by MST
> > > sinks during the disable/enable sequence, those were causing source
> > > to try to do a link training while it was not ready causing CPU
> > > pipe
> > > FIFO underrrus on TGL.
> > > 
> > > v2: Only unsetting TGL_TRANS_DDI_PORT_MASK for TGL on the post
> > > disable sequence
> > > 
> > > v4: Rebased, moved MST sequences to intel_mst_post_disable_dp()
> > > 
> > > BSpec: 4231
> > > BSpec: 4163
> > > BSpec: 22243
> > > BSpec: 49190
> > > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > > Cc: Lucas De Marchi <lucas.demarchi@intel.com>
> > > Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/display/intel_ddi.c    | 33 +++++++++++++++
> > > ------
> > >  drivers/gpu/drm/i915/display/intel_dp_mst.c | 33 +++++++++++++--
> > > ------
> > >  2 files changed, 44 insertions(+), 22 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c
> > > b/drivers/gpu/drm/i915/display/intel_ddi.c
> > > index 9d99ec82d072..94ca26be2fee 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_ddi.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_ddi.c
> > > @@ -34,6 +34,7 @@
> > >  #include "intel_ddi.h"
> > >  #include "intel_display_types.h"
> > >  #include "intel_dp.h"
> > > +#include "intel_dp_mst.h"
> > >  #include "intel_dp_link_training.h"
> > >  #include "intel_dpio_phy.h"
> > >  #include "intel_dsi.h"
> > > @@ -1949,17 +1950,19 @@ void
> > > intel_ddi_disable_transcoder_func(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;
> > > -	i915_reg_t reg = TRANS_DDI_FUNC_CTL(cpu_transcoder);
> > > -	u32 val = I915_READ(reg);
> > > +	u32 val;
> > > +
> > > +	val = I915_READ(TRANS_DDI_FUNC_CTL(cpu_transcoder));
> > > +	val &= ~TRANS_DDI_FUNC_ENABLE;
> > >  
> > >  	if (INTEL_GEN(dev_priv) >= 12) {
> > > -		val &= ~(TRANS_DDI_FUNC_ENABLE |
> > > TGL_TRANS_DDI_PORT_MASK |
> > > -			 TRANS_DDI_DP_VC_PAYLOAD_ALLOC);
> > > +		if (!intel_crtc_has_type(crtc_state,
> > > INTEL_OUTPUT_DP_MST) ||
> > > +		    intel_dp_mst_is_slave_trans(crtc_state))
> > 
> > if (!intel_dp_mst_is_master_trans())
> > 	val &= ...;
> 
> Makes cleaner, thanks
> 
> > 
> > ?
> > 
> > > +			val &= ~TGL_TRANS_DDI_PORT_MASK;
> > >  	} else {
> > > -		val &= ~(TRANS_DDI_FUNC_ENABLE | TRANS_DDI_PORT_MASK |
> > > -			 TRANS_DDI_DP_VC_PAYLOAD_ALLOC);
> > > +		val &= ~TRANS_DDI_PORT_MASK;
> > >  	}
> > > -	I915_WRITE(reg, val);
> > > +	I915_WRITE(TRANS_DDI_FUNC_CTL(cpu_transcoder), val);
> > >  
> > >  	if (dev_priv->quirks & QUIRK_INCREASE_DDI_DISABLED_TIME &&
> > >  	    intel_crtc_has_type(crtc_state, INTEL_OUTPUT_HDMI)) {
> > > @@ -3808,8 +3811,20 @@ static void intel_ddi_post_disable_dp(struct
> > > intel_encoder *encoder,
> > >  	 */
> > >  	intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_OFF);
> > >  
> > > -	if (INTEL_GEN(dev_priv) < 12 && !is_mst)
> > > -		intel_ddi_disable_pipe_clock(old_crtc_state);
> > > +	if (INTEL_GEN(dev_priv) >= 12) {
> > > +		if (is_mst) {
> > 
> > if (intel_dp_mst_is_master_trans()) {
> 
> If you are talking about replace "if (is_mst)" by "if
> (intel_dp_mst_is_master_trans())" both are the same as this function
> will only be called for master. But if you are thinking in replace "if
> (INTEL_GEN(dev_priv) >= 12)" too it will not work.
> 
> Will keep is_mst as we already have it the stack of this functions and
> is shorter

Was just trying to match against the spec. is_master_trans() is
1:1 with the spec so least amount of thinking required.

> 
> > 
> > ?
> > 
> > > +			enum transcoder cpu_transcoder;
> > > +			u32 val;
> > > +
> > > +			cpu_transcoder = old_crtc_state-
> > > >cpu_transcoder;
> > 
> > Assignment can be done when declaring the variable.
> 
> Okay it is less than 100col.
> 
> > 
> > > +			val =
> > > I915_READ(TRANS_DDI_FUNC_CTL(cpu_transcoder));
> > > +			val &= ~TGL_TRANS_DDI_PORT_MASK;
> > > +			I915_WRITE(TRANS_DDI_FUNC_CTL(cpu_transcoder),
> > > val);
> > > +		}
> > > +	} else {
> > > +		if (!is_mst)
> > > +			intel_ddi_disable_pipe_clock(old_crtc_state);
> > > +	}
> > >  
> > >  	intel_disable_ddi_buf(encoder, old_crtc_state);
> > >  
> > > diff --git a/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > b/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > index 710137984c71..efd14b0b507b 100644
> > > --- a/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > +++ b/drivers/gpu/drm/i915/display/intel_dp_mst.c
> > > @@ -347,6 +347,7 @@ static void intel_mst_post_disable_dp(struct
> > > intel_encoder *encoder,
> > >  		to_intel_connector(old_conn_state->connector);
> > >  	struct drm_i915_private *dev_priv = to_i915(connector-
> > > >base.dev);
> > >  	bool last_mst_stream;
> > > +	u32 val;
> > >  
> > >  	intel_dp->active_mst_links--;
> > >  	last_mst_stream = intel_dp->active_mst_links == 0;
> > > @@ -357,6 +358,19 @@ static void intel_mst_post_disable_dp(struct
> > > intel_encoder *encoder,
> > >  
> > >  	intel_disable_pipe(old_crtc_state);
> > >  
> > > +	drm_dp_update_payload_part2(&intel_dp->mst_mgr);
> > 
> > Hmm. I'm having hard to figuring out what these things do. But from a
> > cursory glance it almost looks like part1 is the one that does the
> > AUX
> > stuff to deallocate stuff. So feels like even that part should be
> > here.
> 
> part2 also do aux stuff.
> I also did not deep dive into this part but other drivers also split
> part1 and part2. 
> 
> > 
> > > +
> > > +	val = I915_READ(TRANS_DDI_FUNC_CTL(old_crtc_state-
> > > >cpu_transcoder));
> > > +	val &= ~TRANS_DDI_DP_VC_PAYLOAD_ALLOC;
> > > +	I915_WRITE(TRANS_DDI_FUNC_CTL(old_crtc_state->cpu_transcoder),
> > > val);
> > > +
> > > +	if (intel_de_wait_for_set(dev_priv, intel_dp-
> > > >regs.dp_tp_status,
> > > +				  DP_TP_STATUS_ACT_SENT, 1))
> > > +		DRM_ERROR("Timed out waiting for ACT sent when
> > > disabling\n");
> > 
> > I guess we were missing this step entirely. Dunno if we want ERROR
> > for
> > this. Not sure how much noise it'll generate when someone forcefully
> > yanks the cable...
> 
> This waits the HW send the ACT, it do not wait for any sink
> confirmation. We dont get any warning when removing the cable.

Not really sure what it waits for. Some internal state machine I guess.
Just wasn't sure that wasn't tied to some actual wiggling of some pins
which would require the other end of the cable to be awake.

> 
> > 
> > > +	drm_dp_check_act_status(&intel_dp->mst_mgr);
> > > +
> > > +	drm_dp_mst_deallocate_vcpi(&intel_dp->mst_mgr, connector-
> > > >port);
> > 
> > This seems to be pure sw stuff so not so critical where it is.
> > Keeping
> > it next to the hw payload deallocation seems like a good approach.
> > 
> > > +
> > >  	intel_ddi_disable_transcoder_func(old_crtc_state);
> > >  
> > >  	if (INTEL_GEN(dev_priv) >= 9)
> > > @@ -364,6 +378,12 @@ static void intel_mst_post_disable_dp(struct
> > > intel_encoder *encoder,
> > >  	else
> > >  		ironlake_pfit_disable(old_crtc_state);
> > >  
> > > +	/*
> > > +	 * Power down mst path before disabling the port, otherwise we
> > > end
> > > +	 * up getting interrupts from the sink upon detecting link
> > > loss.
> > > +	 */
> > > +	drm_dp_send_power_updown_phy(&intel_dp->mst_mgr, connector-
> > > >port,
> > > +				     false);
> > 
> > And I guess this was needed to avoid the link loss screams from the
> > sink?
> 
> Yes, we were doing it too late and sink was not happy.
> 
> > 
> > Seems to match the spec better than before at least. Still a bit
> > unsure
> > about he part1/2 stuff for the deallocation. But if that should be
> > changed we can do it as a followup.
> 
> Sure
> 
> > 
> > The code movement into .post_disable() really did make for a much
> > nicer experience when comparing with the spec. Or at least I didn't
> > lose as much hair as with the previous version :)
> 
> Thanks for that, I had imagined a way bigger change when you suggested
> it.
> 
> > 
> > Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > 
> > >  	/*
> > >  	 * From TGL spec: "If multi-stream slave transcoder: Configure
> > >  	 * Transcoder Clock Select to direct no clock to the
> > > transcoder"
> > > @@ -374,19 +394,6 @@ static void intel_mst_post_disable_dp(struct
> > > intel_encoder *encoder,
> > >  	if (INTEL_GEN(dev_priv) < 12 || !last_mst_stream)
> > >  		intel_ddi_disable_pipe_clock(old_crtc_state);
> > >  
> > > -	/* this can fail */
> > > -	drm_dp_check_act_status(&intel_dp->mst_mgr);
> > > -	/* and this can also fail */
> > > -	drm_dp_update_payload_part2(&intel_dp->mst_mgr);
> > > -
> > > -	drm_dp_mst_deallocate_vcpi(&intel_dp->mst_mgr, connector-
> > > >port);
> > > -
> > > -	/*
> > > -	 * Power down mst path before disabling the port, otherwise we
> > > end
> > > -	 * up getting interrupts from the sink upon detecting link
> > > loss.
> > > -	 */
> > > -	drm_dp_send_power_updown_phy(&intel_dp->mst_mgr, connector-
> > > >port,
> > > -				     false);
> > >  
> > >  	intel_mst->connector = NULL;
> > >  	if (last_mst_stream)
> > > -- 
> > > 2.24.1

-- 
Ville Syrjälä
Intel
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx