drm/i915: Implement ReadHitWriteOnlyDisable.
authorRafael Antognolli <rafael.antognolli@intel.com>
Fri, 3 Nov 2017 18:30:27 +0000 (11:30 -0700)
committerRodrigo Vivi <rodrigo.vivi@intel.com>
Wed, 8 Nov 2017 20:43:17 +0000 (12:43 -0800)
The workaround for this is described as:

"if RenderSurfaceState.Num_Multisamples > 1, disable RCC clock gating if
RenderSurfaceState.Num_Multisamples == 1, set 0x7010[14] = 1"

Further documentation in the internal bug referenced by the bspec
suggest that any of the above suggestions should suffice to fix the
issue. We are going with disabling RCC clock gating.

Unfortunately, what we are doing doesn't match the name of the
workaround, but at least it matches its description.

This change improves CNL stability by avoiding some of the hangs seen in
the platform.

v2: Only disable RCC clock gating.

Signed-off-by: Rafael Antognolli <rafael.antognolli@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171103183027.5051-1-rafael.antognolli@intel.com
drivers/gpu/drm/i915/i915_reg.h
drivers/gpu/drm/i915/intel_engine_cs.c

index f0f8f60..6ef3342 100644 (file)
@@ -3837,6 +3837,7 @@ enum {
  */
 #define SLICE_UNIT_LEVEL_CLKGATE       _MMIO(0x94d4)
 #define  SARBUNIT_CLKGATE_DIS          (1 << 5)
+#define  RCCUNIT_CLKGATE_DIS           (1 << 7)
 
 /*
  * Display engine regs
index 6997306..6cb8e3e 100644 (file)
@@ -1320,6 +1320,9 @@ static int cnl_init_workarounds(struct intel_engine_cs *engine)
        WA_SET_FIELD_MASKED(GEN8_CS_CHICKEN1, GEN9_PREEMPT_GPGPU_LEVEL_MASK,
                            GEN9_PREEMPT_GPGPU_COMMAND_LEVEL);
 
+       /* ReadHitWriteOnlyDisable: cnl */
+       WA_SET_BIT_MASKED(SLICE_UNIT_LEVEL_CLKGATE, RCCUNIT_CLKGATE_DIS);
+
        /* WaEnablePreemptionGranularityControlByUMD:cnl */
        I915_WRITE(GEN7_FF_SLICE_CS_CHICKEN1,
                   _MASKED_BIT_ENABLE(GEN9_FFSC_PERCTX_PREEMPT_CTRL));