drm/radeon: Only flush HDP cache for indirect buffers from userspace

It isn't necessary for command streams generated by the kernel (at least
not while we aren't storing ring or indirect buffers in VRAM).

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/radeon/cik.c

@@ -3801,7 +3801,7 @@ int cik_ring_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	radeon_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
 	radeon_ring_write(ring, ((scratch - PACKET3_SET_UCONFIG_REG_START) >> 2));
 	radeon_ring_write(ring, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		tmp = RREG32(scratch);
@@ -4004,7 +4004,7 @@ int cik_copy_cpdma(struct radeon_device *rdev,
 		return r;
 	}
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	radeon_semaphore_free(rdev, &sem, *fence);
 
 	return r;
@@ -4103,7 +4103,7 @@ int cik_ib_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START) >> 2);
 	ib.ptr[2] = 0xDEADBEEF;
 	ib.length_dw = 3;
-	r = radeon_ib_schedule(rdev, &ib, NULL);
+	r = radeon_ib_schedule(rdev, &ib, NULL, false);
 	if (r) {
 		radeon_scratch_free(rdev, scratch);
 		radeon_ib_free(rdev, &ib);
@@ -4324,7 +4324,7 @@ static int cik_cp_gfx_start(struct radeon_device *rdev)
 	radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
 	radeon_ring_write(ring, 0x00000010); /* VGT_OUT_DEALLOC_CNTL */
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	return 0;
 }

drivers/gpu/drm/radeon/cik_sdma.c

@@ -596,7 +596,7 @@ int cik_copy_dma(struct radeon_device *rdev,
 		return r;
 	}
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	radeon_semaphore_free(rdev, &sem, *fence);
 
 	return r;
@@ -638,7 +638,7 @@ int cik_sdma_ring_test(struct radeon_device *rdev,
 	radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr));
 	radeon_ring_write(ring, 1); /* number of DWs to follow */
 	radeon_ring_write(ring, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		tmp = readl(ptr);
@@ -695,7 +695,7 @@ int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	ib.ptr[4] = 0xDEADBEEF;
 	ib.length_dw = 5;
 
-	r = radeon_ib_schedule(rdev, &ib, NULL);
+	r = radeon_ib_schedule(rdev, &ib, NULL, false);
 	if (r) {
 		radeon_ib_free(rdev, &ib);
 		DRM_ERROR("radeon: failed to schedule ib (%d).\n", r);

drivers/gpu/drm/radeon/evergreen.c

@@ -2869,7 +2869,7 @@ static int evergreen_cp_start(struct radeon_device *rdev)
 	radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
 	radeon_ring_write(ring, 0);
 	radeon_ring_write(ring, 0);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	cp_me = 0xff;
 	WREG32(CP_ME_CNTL, cp_me);
@@ -2912,7 +2912,7 @@ static int evergreen_cp_start(struct radeon_device *rdev)
 	radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
 	radeon_ring_write(ring, 0x00000010); /*  */
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	return 0;
 }

drivers/gpu/drm/radeon/evergreen_dma.c

@@ -155,7 +155,7 @@ int evergreen_copy_dma(struct radeon_device *rdev,
 		return r;
 	}
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	radeon_semaphore_free(rdev, &sem, *fence);
 
 	return r;

drivers/gpu/drm/radeon/ni.c

@@ -1505,7 +1505,7 @@ static int cayman_cp_start(struct radeon_device *rdev)
 	radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
 	radeon_ring_write(ring, 0);
 	radeon_ring_write(ring, 0);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	cayman_cp_enable(rdev, true);
 
@@ -1547,7 +1547,7 @@ static int cayman_cp_start(struct radeon_device *rdev)
 	radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
 	radeon_ring_write(ring, 0x00000010); /*  */
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	/* XXX init other rings */
 

drivers/gpu/drm/radeon/r100.c

@@ -925,7 +925,7 @@ int r100_copy_blit(struct radeon_device *rdev,
 	if (fence) {
 		r = radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX);
 	}
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	return r;
 }
 
@@ -958,7 +958,7 @@ void r100_ring_start(struct radeon_device *rdev, struct radeon_ring *ring)
 			  RADEON_ISYNC_ANY3D_IDLE2D |
 			  RADEON_ISYNC_WAIT_IDLEGUI |
 			  RADEON_ISYNC_CPSCRATCH_IDLEGUI);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 }
 
 
@@ -3638,7 +3638,7 @@ int r100_ring_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	}
 	radeon_ring_write(ring, PACKET0(scratch, 0));
 	radeon_ring_write(ring, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		tmp = RREG32(scratch);
 		if (tmp == 0xDEADBEEF) {
@@ -3700,7 +3700,7 @@ int r100_ib_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	ib.ptr[6] = PACKET2(0);
 	ib.ptr[7] = PACKET2(0);
 	ib.length_dw = 8;
-	r = radeon_ib_schedule(rdev, &ib, NULL);
+	r = radeon_ib_schedule(rdev, &ib, NULL, false);
 	if (r) {
 		DRM_ERROR("radeon: failed to schedule ib (%d).\n", r);
 		goto free_ib;

drivers/gpu/drm/radeon/r200.c

@@ -121,7 +121,7 @@ int r200_copy_dma(struct radeon_device *rdev,
 	if (fence) {
 		r = radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX);
 	}
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	return r;
 }
 

drivers/gpu/drm/radeon/r300.c

@@ -295,7 +295,7 @@ void r300_ring_start(struct radeon_device *rdev, struct radeon_ring *ring)
 	radeon_ring_write(ring,
 			  R300_GEOMETRY_ROUND_NEAREST |
 			  R300_COLOR_ROUND_NEAREST);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 }
 
 static void r300_errata(struct radeon_device *rdev)

drivers/gpu/drm/radeon/r420.c

@@ -219,7 +219,7 @@ static void r420_cp_errata_init(struct radeon_device *rdev)
 	radeon_ring_write(ring, PACKET0(R300_CP_RESYNC_ADDR, 1));
 	radeon_ring_write(ring, rdev->config.r300.resync_scratch);
 	radeon_ring_write(ring, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 }
 
 static void r420_cp_errata_fini(struct radeon_device *rdev)
@@ -232,7 +232,7 @@ static void r420_cp_errata_fini(struct radeon_device *rdev)
 	radeon_ring_lock(rdev, ring, 8);
 	radeon_ring_write(ring, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
 	radeon_ring_write(ring, R300_RB3D_DC_FINISH);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	radeon_scratch_free(rdev, rdev->config.r300.resync_scratch);
 }
 

drivers/gpu/drm/radeon/r600.c

@@ -2547,7 +2547,7 @@ int r600_cp_start(struct radeon_device *rdev)
 	radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
 	radeon_ring_write(ring, 0);
 	radeon_ring_write(ring, 0);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	cp_me = 0xff;
 	WREG32(R_0086D8_CP_ME_CNTL, cp_me);
@@ -2683,7 +2683,7 @@ int r600_ring_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
 	radeon_ring_write(ring, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2));
 	radeon_ring_write(ring, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		tmp = RREG32(scratch);
 		if (tmp == 0xDEADBEEF)
@@ -2845,7 +2845,7 @@ int r600_copy_cpdma(struct radeon_device *rdev,
 		return r;
 	}
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	radeon_semaphore_free(rdev, &sem, *fence);
 
 	return r;
@@ -3165,7 +3165,7 @@ int r600_ib_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	ib.ptr[1] = ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2);
 	ib.ptr[2] = 0xDEADBEEF;
 	ib.length_dw = 3;
-	r = radeon_ib_schedule(rdev, &ib, NULL);
+	r = radeon_ib_schedule(rdev, &ib, NULL, false);
 	if (r) {
 		DRM_ERROR("radeon: failed to schedule ib (%d).\n", r);
 		goto free_ib;

drivers/gpu/drm/radeon/r600_dma.c

@@ -261,7 +261,7 @@ int r600_dma_ring_test(struct radeon_device *rdev,
 	radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc);
 	radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xff);
 	radeon_ring_write(ring, 0xDEADBEEF);
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 
 	for (i = 0; i < rdev->usec_timeout; i++) {
 		tmp = readl(ptr);
@@ -368,7 +368,7 @@ int r600_dma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring)
 	ib.ptr[3] = 0xDEADBEEF;
 	ib.length_dw = 4;
 
-	r = radeon_ib_schedule(rdev, &ib, NULL);
+	r = radeon_ib_schedule(rdev, &ib, NULL, false);
 	if (r) {
 		radeon_ib_free(rdev, &ib);
 		DRM_ERROR("radeon: failed to schedule ib (%d).\n", r);
@@ -493,7 +493,7 @@ int r600_copy_dma(struct radeon_device *rdev,
 		return r;
 	}
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	radeon_semaphore_free(rdev, &sem, *fence);
 
 	return r;

drivers/gpu/drm/radeon/radeon.h

@@ -968,7 +968,7 @@ int radeon_ib_get(struct radeon_device *rdev, int ring,
 		  unsigned size);
 void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib *ib);
 int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib,
-		       struct radeon_ib *const_ib);
+		       struct radeon_ib *const_ib, bool hdp_flush);
 int radeon_ib_pool_init(struct radeon_device *rdev);
 void radeon_ib_pool_fini(struct radeon_device *rdev);
 int radeon_ib_ring_tests(struct radeon_device *rdev);
@@ -978,8 +978,10 @@ bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev,
 void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *cp);
 int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw);
 int radeon_ring_lock(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw);
-void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *cp);
-void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_ring *cp);
+void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *cp,
+			bool hdp_flush);
+void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_ring *cp,
+			       bool hdp_flush);
 void radeon_ring_undo(struct radeon_ring *ring);
 void radeon_ring_unlock_undo(struct radeon_device *rdev, struct radeon_ring *cp);
 int radeon_ring_test(struct radeon_device *rdev, struct radeon_ring *cp);

drivers/gpu/drm/radeon/radeon_cs.c

@@ -451,7 +451,7 @@ static int radeon_cs_ib_chunk(struct radeon_device *rdev,
 		radeon_vce_note_usage(rdev);
 
 	radeon_cs_sync_rings(parser);
-	r = radeon_ib_schedule(rdev, &parser->ib, NULL);
+	r = radeon_ib_schedule(rdev, &parser->ib, NULL, true);
 	if (r) {
 		DRM_ERROR("Failed to schedule IB !\n");
 	}
@@ -542,9 +542,9 @@ static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev,
 
 	if ((rdev->family >= CHIP_TAHITI) &&
 	    (parser->chunk_const_ib_idx != -1)) {
-		r = radeon_ib_schedule(rdev, &parser->ib, &parser->const_ib);
+		r = radeon_ib_schedule(rdev, &parser->ib, &parser->const_ib, true);
 	} else {
-		r = radeon_ib_schedule(rdev, &parser->ib, NULL);
+		r = radeon_ib_schedule(rdev, &parser->ib, NULL, true);
 	}
 
 out:

drivers/gpu/drm/radeon/radeon_ib.c

@@ -107,6 +107,7 @@ void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib *ib)
  * @rdev: radeon_device pointer
  * @ib: IB object to schedule
  * @const_ib: Const IB to schedule (SI only)
+ * @hdp_flush: Whether or not to perform an HDP cache flush
  *
  * Schedule an IB on the associated ring (all asics).
  * Returns 0 on success, error on failure.
@@ -122,7 +123,7 @@ void radeon_ib_free(struct radeon_device *rdev, struct radeon_ib *ib)
  * to SI there was just a DE IB.
  */
 int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib,
-		       struct radeon_ib *const_ib)
+		       struct radeon_ib *const_ib, bool hdp_flush)
 {
 	struct radeon_ring *ring = &rdev->ring[ib->ring];
 	int r = 0;
@@ -176,7 +177,7 @@ int radeon_ib_schedule(struct radeon_device *rdev, struct radeon_ib *ib,
 	if (ib->vm)
 		radeon_vm_fence(rdev, ib->vm, ib->fence);
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, hdp_flush);
 	return 0;
 }
 

drivers/gpu/drm/radeon/radeon_ring.c

@@ -177,16 +177,18 @@ int radeon_ring_lock(struct radeon_device *rdev, struct radeon_ring *ring, unsig
  *
  * @rdev: radeon_device pointer
  * @ring: radeon_ring structure holding ring information
+ * @hdp_flush: Whether or not to perform an HDP cache flush
  *
  * Update the wptr (write pointer) to tell the GPU to
  * execute new commands on the ring buffer (all asics).
  */
-void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *ring)
+void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *ring,
+			bool hdp_flush)
 {
 	/* If we are emitting the HDP flush via the ring buffer, we need to
 	 * do it before padding.
 	 */
-	if (rdev->asic->ring[ring->idx]->hdp_flush)
+	if (hdp_flush && rdev->asic->ring[ring->idx]->hdp_flush)
 		rdev->asic->ring[ring->idx]->hdp_flush(rdev, ring);
 	/* We pad to match fetch size */
 	while (ring->wptr & ring->align_mask) {
@@ -196,7 +198,7 @@ void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *ring)
 	/* If we are emitting the HDP flush via MMIO, we need to do it after
 	 * all CPU writes to VRAM finished.
 	 */
-	if (rdev->asic->mmio_hdp_flush)
+	if (hdp_flush && rdev->asic->mmio_hdp_flush)
 		rdev->asic->mmio_hdp_flush(rdev);
 	radeon_ring_set_wptr(rdev, ring);
 }
@@ -207,12 +209,14 @@ void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *ring)
  *
  * @rdev: radeon_device pointer
  * @ring: radeon_ring structure holding ring information
+ * @hdp_flush: Whether or not to perform an HDP cache flush
  *
  * Call radeon_ring_commit() then unlock the ring (all asics).
  */
-void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_ring *ring)
+void radeon_ring_unlock_commit(struct radeon_device *rdev, struct radeon_ring *ring,
+			       bool hdp_flush)
 {
-	radeon_ring_commit(rdev, ring);
+	radeon_ring_commit(rdev, ring, hdp_flush);
 	mutex_unlock(&rdev->ring_lock);
 }
 
@@ -372,7 +376,7 @@ int radeon_ring_restore(struct radeon_device *rdev, struct radeon_ring *ring,
 		radeon_ring_write(ring, data[i]);
 	}
 
-	radeon_ring_unlock_commit(rdev, ring);
+	radeon_ring_unlock_commit(rdev, ring, false);
 	kfree(data);
 	return 0;
 }

drivers/gpu/drm/radeon/radeon_semaphore.c

@@ -179,7 +179,7 @@ int radeon_semaphore_sync_rings(struct radeon_device *rdev,
 			continue;
 		}
 
-		radeon_ring_commit(rdev, &rdev->ring[i]);
+		radeon_ring_commit(rdev, &rdev->ring[i], false);
 		radeon_fence_note_sync(fence, ring);
 
 		semaphore->gpu_addr += 8;