drm/amdgpu: modify packet size for pm4 flush tlbs

[Why]
PM4 packet size for flush message was oversized.

[How]
Packet size adjusted to allocate flush + fence packets.
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c

@@ -368,7 +368,7 @@ static const struct kiq_pm4_funcs gfx_v10_0_kiq_pm4_funcs = {
 	.map_queues_size = 7,
 	.unmap_queues_size = 6,
 	.query_status_size = 7,
-	.invalidate_tlbs_size = 12,
+	.invalidate_tlbs_size = 2,
 };
 
 static void gfx_v10_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -859,7 +859,7 @@ static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = {
 	.map_queues_size = 7,
 	.unmap_queues_size = 6,
 	.query_status_size = 7,
-	.invalidate_tlbs_size = 12,
+	.invalidate_tlbs_size = 2,
 };
 
 static void gfx_v9_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c

@@ -418,7 +418,8 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
 
 	if (amdgpu_emu_mode == 0 && ring->sched.ready) {
 		spin_lock(&adev->gfx.kiq.ring_lock);
-		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size);
+		/* 2 dwords flush + 8 dwords fence */
+		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8);
 		kiq->pmf->kiq_invalidate_tlbs(ring,
 					pasid, flush_type, all_hub);
 		amdgpu_fence_emit_polling(ring, &seq);

drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c

@@ -578,7 +578,8 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
 
 	if (ring->sched.ready) {
 		spin_lock(&adev->gfx.kiq.ring_lock);
-		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size);
+		/* 2 dwords flush + 8 dwords fence */
+		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8);
 		kiq->pmf->kiq_invalidate_tlbs(ring,
 					pasid, flush_type, all_hub);
 		amdgpu_fence_emit_polling(ring, &seq);