[Intel-gfx] [PATCH] drm/i915/tgl: Implement Wa_1606931601

[Matt Roper]

On Wed, Jan 29, 2020 at 02:42:06PM -0800, Anusha Srivatsa wrote:
> Disable Inter and intra Read Suppression (bit 15) and
> Early Read and Src Swap (bit 14) by setting the chicken
> register.
> 
> BSpec: 46045,52890
> 
> v2: Follow the Bspec implementation for the WA.
> v3: Have 2 separate defines for bit 14 and 15.
> - Rename register definitions with TGL_ prefix

The hardware guys changed their mind again and we're back to only
needing bit 14 now.  They updated the bspec and the underlying database
yet again.  :-/

> 
> Cc: Matt Roper <matthew.d.roper at intel.com>
> Signed-off-by: Anusha Srivatsa <anusha.srivatsa at intel.com>
> ---
>  drivers/gpu/drm/i915/gt/intel_workarounds.c | 6 ++++++
>  drivers/gpu/drm/i915/i915_reg.h             | 2 ++
>  2 files changed, 8 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> index 5a7db279f702..1f84cd595f88 100644
> --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> @@ -593,6 +593,12 @@ static void tgl_ctx_workarounds_init(struct intel_engine_cs *engine,
>  	wa_add(wal, FF_MODE2, FF_MODE2_TDS_TIMER_MASK, val,
>  	       IS_TGL_REVID(engine->i915, TGL_REVID_A0, TGL_REVID_A0) ? 0 :
>  			    FF_MODE2_TDS_TIMER_MASK);
> +
> +	/* Wa_1606931601:tgl */
> +	WA_SET_BIT_MASKED(GEN7_ROW_CHICKEN2,
> +			  GEN12_EARLY_READ_SRC0_DISABLE |
> +			  GEN12_INTER_INTRA_READ_SUPPRESSION_DISABLE);

I think Daniele already mentioned this on the other discussion, but
ROW_CHICKEN2 isn't part of the context image on gen12 (see bspec page
46255).  This is a change from ICL where it *was* part of the context
(see bspec page 18907), so even though we handled this register in the
ctx_workarounds_init for ICL, that's not the appropriate place to put it
for TGL.

Since this isn't a context workaround, we need to determine whether it's
a general GT workaround (which would be initialized in
tgl_gt_workarounds_init) or an engine workaround (which would be
initialized in {rcs,xcs}_engine_wa_init.  In this case the register
we're modifying is 0xe49c; according to bspec page 52078 this falls in
one of the forcewake ranges listed under the render engine column
(0E000-0E8FF).  So I believe in this case that means we want to update
rcs_engine_wa_init() with this workaround --- that ensures that the
workaround will be re-applied any time the engine is reset (even if it's
not a full-GPU reset).



Matt

> +
>  }
>  
>  static void
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 4c72b8ac0f2e..70ead809c706 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -9149,6 +9149,8 @@ enum {
>  #define   DOP_CLOCK_GATING_DISABLE	(1 << 0)
>  #define   PUSH_CONSTANT_DEREF_DISABLE	(1 << 8)
>  #define   GEN11_TDL_CLOCK_GATING_FIX_DISABLE	(1 << 1)
> +#define   GEN12_EARLY_READ_SRC0_DISABLE		(1 << 14)
> +#define   GEN12_INTER_INTRA_READ_SUPPRESSION_DISABLE	(1 << 15)
>  
>  #define HSW_ROW_CHICKEN3		_MMIO(0xe49c)
>  #define  HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE    (1 << 6)
> -- 
> 2.25.0
> 

-- 
Matt Roper
Graphics Software Engineer
VTT-OSGC Platform Enablement
Intel Corporation
(916) 356-2795