[PATCH v3] drm/i915/vlv: Added a rendering specific Hw WA 'WaTlbInvalidateStoreDataBefore'

[PATCH v3] drm/i915/vlv: Added a rendering specific Hw WA 'WaTlbInvalidateStoreDataBefore'

[sourab.gupta]

From: Akash Goel <akash.goel@intel.com>

Added a new rendering specific Workaround 'WaTlbInvalidateStoreDataBefore'.
In this WA, before pipecontrol with TLB invalidate set, need to add 2 MI
Store data commands.

v2: Modified the WA comment (Ville)

v3: Added the vlv identifier with WA name (Damien)

Signed-off-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Sourab Gupta <sourab.gupta@intel.com>
---
 drivers/gpu/drm/i915/intel_ringbuffer.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 4eb3e06..2cc7ed5 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2207,6 +2207,28 @@ intel_ring_invalidate_all_caches(struct intel_ring_buffer *ring)
 	uint32_t flush_domains;
 	int ret;
 
+	if (IS_VALLEYVIEW(ring->dev)) {
+		/*
+		 * WaTlbInvalidateStoreDataBefore:vlv
+		 * Before pipecontrol with TLB invalidate set, need 2 store
+		 * data commands (such as MI_STORE_DATA_IMM or MI_STORE_DATA_INDEX)
+		 * Without this, hardware cannot guarantee the command after the
+		 * PIPE_CONTROL with TLB inv will not use the old TLB values.
+		 */
+		int i;
+		ret = intel_ring_begin(ring, 4 * 2);
+		if (ret)
+			return ret;
+		for (i = 0; i < 2; i++) {
+			intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
+			intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_INDEX <<
+						MI_STORE_DWORD_INDEX_SHIFT);
+			intel_ring_emit(ring, 0);
+			intel_ring_emit(ring, MI_NOOP);
+		}
+		intel_ring_advance(ring);
+	}
+
 	flush_domains = 0;
 	if (ring->gpu_caches_dirty)
 		flush_domains = I915_GEM_GPU_DOMAINS;
-- 
1.8.5.1

Re: [PATCH v3] drm/i915/vlv: Added a rendering specific Hw WA 'WaTlbInvalidateStoreDataBefore'

[Daniel Vetter]

On Fri, Mar 21, 2014 at 06:02:36PM +0530, sourab.gupta@intel.com wrote:
> From: Akash Goel <akash.goel@intel.com>
> 
> Added a new rendering specific Workaround 'WaTlbInvalidateStoreDataBefore'.
> In this WA, before pipecontrol with TLB invalidate set, need to add 2 MI
> Store data commands.
> 
> v2: Modified the WA comment (Ville)
> 
> v3: Added the vlv identifier with WA name (Damien)
> 
> Signed-off-by: Akash Goel <akash.goel@intel.com>
> Signed-off-by: Sourab Gupta <sourab.gupta@intel.com>

We seem to have a bit a mess of vlv w/a patches with them splattered over
multiple threads and the discussion split up equally badly. And I'm no
longer clear at all which patches are superseeded and where exactly I can
find the latest version of each.

Can you please dig up all the vlv w/a patches you have in-flight and
resubmit them in a new thread?

Also when resubmitting individual patches of a series you need to set the
in-reply-to for each individual patch to it's previous version, otherwise
it's really hard to follow. Top-posting and lumping a few resends together
just confuses things.

And if a thread gets too messy the best option is always to start a new
clean thread.
-Daniel

> ---
>  drivers/gpu/drm/i915/intel_ringbuffer.c | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> index 4eb3e06..2cc7ed5 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> @@ -2207,6 +2207,28 @@ intel_ring_invalidate_all_caches(struct intel_ring_buffer *ring)
>  	uint32_t flush_domains;
>  	int ret;
>  
> +	if (IS_VALLEYVIEW(ring->dev)) {
> +		/*
> +		 * WaTlbInvalidateStoreDataBefore:vlv
> +		 * Before pipecontrol with TLB invalidate set, need 2 store
> +		 * data commands (such as MI_STORE_DATA_IMM or MI_STORE_DATA_INDEX)
> +		 * Without this, hardware cannot guarantee the command after the
> +		 * PIPE_CONTROL with TLB inv will not use the old TLB values.
> +		 */
> +		int i;
> +		ret = intel_ring_begin(ring, 4 * 2);
> +		if (ret)
> +			return ret;
> +		for (i = 0; i < 2; i++) {
> +			intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
> +			intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_INDEX <<
> +						MI_STORE_DWORD_INDEX_SHIFT);
> +			intel_ring_emit(ring, 0);
> +			intel_ring_emit(ring, MI_NOOP);
> +		}
> +		intel_ring_advance(ring);
> +	}
> +
>  	flush_domains = 0;
>  	if (ring->gpu_caches_dirty)
>  		flush_domains = I915_GEM_GPU_DOMAINS;
> -- 
> 1.8.5.1
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Re: [PATCH v3] drm/i915/vlv: Added a rendering specific Hw WA 'WaTlbInvalidateStoreDataBefore'

[Gupta, Sourab]

On Fri, 2014-03-21 at 15:15 +0000, Daniel Vetter wrote:
> On Fri, Mar 21, 2014 at 06:02:36PM +0530, sourab.gupta@intel.com wrote:
> > From: Akash Goel <akash.goel@intel.com>
> > 
> > Added a new rendering specific Workaround 'WaTlbInvalidateStoreDataBefore'.
> > In this WA, before pipecontrol with TLB invalidate set, need to add 2 MI
> > Store data commands.
> > 
> > v2: Modified the WA comment (Ville)
> > 
> > v3: Added the vlv identifier with WA name (Damien)
> > 
> > Signed-off-by: Akash Goel <akash.goel@intel.com>
> > Signed-off-by: Sourab Gupta <sourab.gupta@intel.com>
> 
> We seem to have a bit a mess of vlv w/a patches with them splattered over
> multiple threads and the discussion split up equally badly. And I'm no
> longer clear at all which patches are superseeded and where exactly I can
> find the latest version of each.
> 
> Can you please dig up all the vlv w/a patches you have in-flight and
> resubmit them in a new thread?
> 
> Also when resubmitting individual patches of a series you need to set the
> in-reply-to for each individual patch to it's previous version, otherwise
> it's really hard to follow. Top-posting and lumping a few resends together
> just confuses things.
> 
> And if a thread gets too messy the best option is always to start a new
> clean thread.
> -Daniel

I agree to your points.
I'll collate the latest versions of the vlv w/a patches and submit them
together shortly.

Regards,
Sourab

> 
> > ---
> >  drivers/gpu/drm/i915/intel_ringbuffer.c | 22 ++++++++++++++++++++++
> >  1 file changed, 22 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> > index 4eb3e06..2cc7ed5 100644
> > --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> > @@ -2207,6 +2207,28 @@ intel_ring_invalidate_all_caches(struct intel_ring_buffer *ring)
> >  	uint32_t flush_domains;
> >  	int ret;
> >  
> > +	if (IS_VALLEYVIEW(ring->dev)) {
> > +		/*
> > +		 * WaTlbInvalidateStoreDataBefore:vlv
> > +		 * Before pipecontrol with TLB invalidate set, need 2 store
> > +		 * data commands (such as MI_STORE_DATA_IMM or MI_STORE_DATA_INDEX)
> > +		 * Without this, hardware cannot guarantee the command after the
> > +		 * PIPE_CONTROL with TLB inv will not use the old TLB values.
> > +		 */
> > +		int i;
> > +		ret = intel_ring_begin(ring, 4 * 2);
> > +		if (ret)
> > +			return ret;
> > +		for (i = 0; i < 2; i++) {
> > +			intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
> > +			intel_ring_emit(ring, I915_GEM_HWS_SCRATCH_INDEX <<
> > +						MI_STORE_DWORD_INDEX_SHIFT);
> > +			intel_ring_emit(ring, 0);
> > +			intel_ring_emit(ring, MI_NOOP);
> > +		}
> > +		intel_ring_advance(ring);
> > +	}
> > +
> >  	flush_domains = 0;
> >  	if (ring->gpu_caches_dirty)
> >  		flush_domains = I915_GEM_GPU_DOMAINS;
> > -- 
> > 1.8.5.1
> > 
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx@lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>