[Intel-gfx] [PATCH 02/15] drm/i915/gtt: Preallocate Braswell top-level page directory

[Chris Wilson]

In order for the Braswell top-level PD to remain the same from the time
of request construction to its submission onto HW, as we may be
asynchronously rewriting the page tables (thus changing the expected
register state after having already stored the old addresses in the
request), the top level PD must be preallocated.

So wave goodbye to our lazy allocation of those 4x2 pages.

v2: A little bit of write-flushing required (presumably it always has
been required, but now we are more susceptible and it is showing up!)

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala at linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala at linux.intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_context.c |  8 ++-
 drivers/gpu/drm/i915/gt/intel_lrc.c         | 59 +--------------------
 drivers/gpu/drm/i915/i915_gem_gtt.c         | 10 ++--
 3 files changed, 13 insertions(+), 64 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index 1cdfe05514c3..863b5cc1ffa9 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -1003,12 +1003,18 @@ static int emit_ppgtt_update(struct i915_request *rq, void *data)
 		intel_ring_advance(rq, cs);
 	} else if (HAS_LOGICAL_RING_CONTEXTS(engine->i915)) {
 		struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(vm);
+		int err;
+
+		/* Magic required to prevent forcewake errors! */
+		err = engine->emit_flush(rq, EMIT_INVALIDATE);
+		if (err)
+			return err;
 
 		cs = intel_ring_begin(rq, 4 * GEN8_3LVL_PDPES + 2);
 		if (IS_ERR(cs))
 			return PTR_ERR(cs);
 
-		*cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES);
+		*cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES) | MI_LRI_FORCE_POSTED;;
 		for (i = GEN8_3LVL_PDPES; i--; ) {
 			const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i);
 
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index d42584439f51..d14569f7637d 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1871,60 +1871,6 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq)
 	return 0;
 }
 
-static int emit_pdps(struct i915_request *rq)
-{
-	const struct intel_engine_cs * const engine = rq->engine;
-	struct i915_ppgtt * const ppgtt = i915_vm_to_ppgtt(rq->hw_context->vm);
-	int err, i;
-	u32 *cs;
-
-	GEM_BUG_ON(intel_vgpu_active(rq->i915));
-
-	/*
-	 * Beware ye of the dragons, this sequence is magic!
-	 *
-	 * Small changes to this sequence can cause anything from
-	 * GPU hangs to forcewake errors and machine lockups!
-	 */
-
-	/* Flush any residual operations from the context load */
-	err = engine->emit_flush(rq, EMIT_FLUSH);
-	if (err)
-		return err;
-
-	/* Magic required to prevent forcewake errors! */
-	err = engine->emit_flush(rq, EMIT_INVALIDATE);
-	if (err)
-		return err;
-
-	cs = intel_ring_begin(rq, 4 * GEN8_3LVL_PDPES + 2);
-	if (IS_ERR(cs))
-		return PTR_ERR(cs);
-
-	/* Ensure the LRI have landed before we invalidate & continue */
-	*cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES) | MI_LRI_FORCE_POSTED;
-	for (i = GEN8_3LVL_PDPES; i--; ) {
-		const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i);
-		u32 base = engine->mmio_base;
-
-		*cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, i));
-		*cs++ = upper_32_bits(pd_daddr);
-		*cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, i));
-		*cs++ = lower_32_bits(pd_daddr);
-	}
-	*cs++ = MI_NOOP;
-
-	intel_ring_advance(rq, cs);
-
-	/* Be doubly sure the LRI have landed before proceeding */
-	err = engine->emit_flush(rq, EMIT_FLUSH);
-	if (err)
-		return err;
-
-	/* Re-invalidate the TLB for luck */
-	return engine->emit_flush(rq, EMIT_INVALIDATE);
-}
-
 static int execlists_request_alloc(struct i915_request *request)
 {
 	int ret;
@@ -1947,10 +1893,7 @@ static int execlists_request_alloc(struct i915_request *request)
 	 */
 
 	/* Unconditionally invalidate GPU caches and TLBs. */
-	if (i915_vm_is_4lvl(request->hw_context->vm))
-		ret = request->engine->emit_flush(request, EMIT_INVALIDATE);
-	else
-		ret = emit_pdps(request);
+	ret = request->engine->emit_flush(request, EMIT_INVALIDATE);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 0b81e0b64393..43d0de15bb21 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -162,6 +162,7 @@ static int ppgtt_bind_vma(struct i915_vma *vma,
 		pte_flags |= PTE_READ_ONLY;
 
 	vma->vm->insert_entries(vma->vm, vma, cache_level, pte_flags);
+	wmb();
 
 	return 0;
 }
@@ -1420,6 +1421,7 @@ static int gen8_preallocate_top_level_pdp(struct i915_ppgtt *ppgtt)
 		set_pd_entry(pd, idx, pde);
 		atomic_inc(px_used(pde)); /* keep pinned */
 	}
+	wmb();
 
 	return 0;
 }
@@ -1507,11 +1509,9 @@ static struct i915_ppgtt *gen8_ppgtt_create(struct drm_i915_private *i915)
 	}
 
 	if (!i915_vm_is_4lvl(&ppgtt->vm)) {
-		if (intel_vgpu_active(i915)) {
-			err = gen8_preallocate_top_level_pdp(ppgtt);
-			if (err)
-				goto err_free_pd;
-		}
+		err = gen8_preallocate_top_level_pdp(ppgtt);
+		if (err)
+			goto err_free_pd;
 	}
 
 	ppgtt->vm.insert_entries = gen8_ppgtt_insert;
-- 
2.23.0