[Intel-gfx] [PATCH] drm/i915: Fix MST disable sequence

* [Intel-gfx] [PATCH] drm/i915: Fix MST disable sequence
@ 2020-01-08 14:24 Ville Syrjala
  2020-01-08 14:45 ` [Intel-gfx] [PATCH v2] " Ville Syrjala
                   ` (2 more replies)
  0 siblings, 3 replies; 8+ messages in thread
From: Ville Syrjala @ 2020-01-08 14:24 UTC (permalink / raw)
  To: intel-gfx

From: Ville Syrjälä <ville.syrjala@linux.intel.com>

When moving the pipe disable & co. function calls from
haswell_crtc_disable() into the encoder .post_disable() hooks I
neglected to account for the MST vs. DDI interactions properly.
This now leads us to call these functions two times for the last
MST stream (once from the MST code and a second time from the DDI
code). The calls from the DDI code should only be done for SST
and not MST. Add the proper check for that.

This results in an MCE on ICL. My vague theory is that we turn off
the transcoder clock from the MST code and then we proceed to touch
something in the DDI code which still depends on that clock causing
the hardware to become upset. Though I can't really explain why
Stan's hack of omitting the pipe disable in the MST code would avoid
the MCE since we should still be turning off the transcoder clock.
But maybe there's something magic in the hw that keeps the clock on
as long as the pipe is on. Or maybe the clock isn't the problem and
we now touch something in the DDI disable code that really does need
the pipe to be still enabled.

Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Manasi Navare <manasi.d.navare@intel.com>
Reported-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Closes: https://gitlab.freedesktop.org/drm/intel/issues/901
Fixes: 773b4b54351c ("drm/i915: Move stuff from haswell_crtc_disable() into encoder .post_disable()")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
---
 drivers/gpu/drm/i915/display/intel_ddi.c | 22 ++++++++++++----------
 1 file changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/display/intel_ddi.c b/drivers/gpu/drm/i915/display/intel_ddi.c
index 730118903608..930e814eb96e 100644
--- a/drivers/gpu/drm/i915/display/intel_ddi.c
+++ b/drivers/gpu/drm/i915/display/intel_ddi.c
@@ -3900,21 +3900,23 @@ static void intel_ddi_post_disable(struct intel_encoder *encoder,
 	enum phy phy = intel_port_to_phy(dev_priv, encoder->port);
 	bool is_tc_port = intel_phy_is_tc(dev_priv, phy);
 
-	intel_crtc_vblank_off(old_crtc_state);
+	if (!intel_crtc_has_type(old_crtc_state, INTEL_OUTPUT_DP_MST)) {
+		intel_crtc_vblank_off(old_crtc_state);
 
-	intel_disable_pipe(old_crtc_state);
+		intel_disable_pipe(old_crtc_state);
 
-	if (INTEL_GEN(dev_priv) >= 11)
-		icl_disable_transcoder_port_sync(old_crtc_state);
+		if (INTEL_GEN(dev_priv) >= 11)
+			icl_disable_transcoder_port_sync(old_crtc_state);
 
-	intel_ddi_disable_transcoder_func(old_crtc_state);
+		intel_ddi_disable_transcoder_func(old_crtc_state);
 
-	intel_dsc_disable(old_crtc_state);
+		intel_dsc_disable(old_crtc_state);
 
-	if (INTEL_GEN(dev_priv) >= 9)
-		skylake_scaler_disable(old_crtc_state);
-	else
-		ironlake_pfit_disable(old_crtc_state);
+		if (INTEL_GEN(dev_priv) >= 9)
+			skylake_scaler_disable(old_crtc_state);
+		else
+			ironlake_pfit_disable(old_crtc_state);
+	}
 
 	/*
 	 * When called from DP MST code:
-- 
2.24.1

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

^ permalink raw reply related	[flat|nested] 8+ messages in thread