Re: [Intel-gfx] [PATCH v5 8/9] drm/i915/dg2: Maintain backward-compatible nested batch behavior

["Srivatsa, Anusha" <anusha.srivatsa@intel.com>]

> -----Original Message-----
> From: Intel-gfx <intel-gfx-bounces@lists.freedesktop.org> On Behalf Of Matt
> Roper
> Sent: Thursday, August 5, 2021 9:37 AM
> To: intel-gfx@lists.freedesktop.org
> Cc: Roper, Matthew D <matthew.d.roper@intel.com>; Harrison, John C
> <john.c.harrison@intel.com>
> Subject: [Intel-gfx] [PATCH v5 8/9] drm/i915/dg2: Maintain backward-
> compatible nested batch behavior
> 
> For tgl+, the per-context setting of MI_MODE[12] determines whether the
> bits of a nested MI_BATCH_BUFFER_START instruction should be interpreted
> in the traditional manner or whether they should instead use a new tgl+
> meaning that breaks backward compatibility, but allows nesting into 3rd-
> level batchbuffers.  For previous platforms, the hardware default for this
> register bit is to maintain backward-compatible behavior unless a context
> intentionally opts into the new behavior; however Xe_HPG flips the hardware
> default behavior.
> 
> From a SW perspective, we want to maintain the backward-compatible
> behavior for userspace, so we'll apply a fake workaround to set it back to the
> legacy behavior on platforms where the hardware default is to break
> compatibility.  At the moment there is no Linux userspace that utilizes third-
> level batchbuffers, so this will avoid userspace from needing to make any
> changes.  using the legacy meaning is the correct thing to do.  If/when we
> have userspace consumers that want to utilize third-level batch nesting, we
> can provide a context parameter to allow them to opt-in.
> 
> Bspec: 45974, 45718
> Cc: John Harrison <John.C.Harrison@Intel.com>
> Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

Looks good.

Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>

> ---
>  drivers/gpu/drm/i915/gt/intel_workarounds.c | 39 +++++++++++++++++++--
>  drivers/gpu/drm/i915/i915_reg.h             |  1 +
>  2 files changed, 38 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> index aae609d7d85d..97b3cd81b721 100644
> --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
> +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
> @@ -644,6 +644,37 @@ static void dg1_ctx_workarounds_init(struct
> intel_engine_cs *engine,
>  		     DG1_HZ_READ_SUPPRESSION_OPTIMIZATION_DISABLE);
>  }
> 
> +static void fakewa_disable_nestedbb_mode(struct intel_engine_cs *engine,
> +					 struct i915_wa_list *wal)
> +{
> +	/*
> +	 * This is a "fake" workaround defined by software to ensure we
> +	 * maintain reliable, backward-compatible behavior for userspace
> with
> +	 * regards to how nested MI_BATCH_BUFFER_START commands are
> handled.
> +	 *
> +	 * The per-context setting of MI_MODE[12] determines whether the
> bits
> +	 * of a nested MI_BATCH_BUFFER_START instruction should be
> interpreted
> +	 * in the traditional manner or whether they should instead use a
> new
> +	 * tgl+ meaning that breaks backward compatibility, but allows
> nesting
> +	 * into 3rd-level batchbuffers.  When this new capability was first
> +	 * added in TGL, it remained off by default unless a context
> +	 * intentionally opted in to the new behavior.  However Xe_HPG now
> +	 * flips this on by default and requires that we explicitly opt out if
> +	 * we don't want the new behavior.
> +	 *
> +	 * From a SW perspective, we want to maintain the backward-
> compatible
> +	 * behavior for userspace, so we'll apply a fake workaround to set it
> +	 * back to the legacy behavior on platforms where the hardware
> default
> +	 * is to break compatibility.  At the moment there is no Linux
> +	 * userspace that utilizes third-level batchbuffers, so this will avoid
> +	 * userspace from needing to make any changes.  using the legacy
> +	 * meaning is the correct thing to do.  If/when we have userspace
> +	 * consumers that want to utilize third-level batch nesting, we can
> +	 * provide a context parameter to allow them to opt-in.
> +	 */
> +	wa_masked_dis(wal, RING_MI_MODE(engine->mmio_base),
> TGL_NESTED_BB_EN);
> +}
> +
>  static void
>  __intel_engine_init_ctx_wa(struct intel_engine_cs *engine,
>  			   struct i915_wa_list *wal,
> @@ -651,11 +682,15 @@ __intel_engine_init_ctx_wa(struct intel_engine_cs
> *engine,  {
>  	struct drm_i915_private *i915 = engine->i915;
> 
> +	wa_init_start(wal, name, engine->name);
> +
> +	/* Applies to all engines */
> +	if (GRAPHICS_VER_FULL(i915) >= IP_VER(12, 55))
> +		fakewa_disable_nestedbb_mode(engine, wal);
> +
>  	if (engine->class != RENDER_CLASS)
>  		return;
> 
> -	wa_init_start(wal, name, engine->name);
> -
>  	if (IS_DG1(i915))
>  		dg1_ctx_workarounds_init(engine, wal);
>  	else if (GRAPHICS_VER(i915) == 12)
> diff --git a/drivers/gpu/drm/i915/i915_reg.h
> b/drivers/gpu/drm/i915/i915_reg.h index 77f6dcaba2b9..269685955fbd
> 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -2821,6 +2821,7 @@ static inline bool i915_mmio_reg_valid(i915_reg_t
> reg)
>  #define MI_MODE		_MMIO(0x209c)
>  # define VS_TIMER_DISPATCH				(1 << 6)
>  # define MI_FLUSH_ENABLE				(1 << 12)
> +# define TGL_NESTED_BB_EN				(1 << 12)
>  # define ASYNC_FLIP_PERF_DISABLE			(1 << 14)
>  # define MODE_IDLE					(1 << 9)
>  # define STOP_RING					(1 << 8)
> --
> 2.25.4