drm/amdgpu: add master/slave check in init phase

Skip KCQ setup on slave xcc as there's no use case.

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 36 insertions, 23 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 7f5c60381103..c83fb4277233 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -489,16 +489,19 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
 		return -EINVAL;
 
 	spin_lock(&kiq->ring_lock);
-	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
-					adev->gfx.num_compute_rings)) {
-		spin_unlock(&adev->gfx.kiq[0].ring_lock);
-		return -ENOMEM;
-	}
+	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+		if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
+						adev->gfx.num_compute_rings)) {
+			spin_unlock(&kiq->ring_lock);
+			return -ENOMEM;
+		}
 
-	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		j = i + xcc_id * adev->gfx.num_compute_rings;
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.compute_ring[i],
-					   RESET_QUEUES, 0, 0);
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_compute_rings;
+			kiq->pmf->kiq_unmap_queues(kiq_ring,
+						   &adev->gfx.compute_ring[i],
+						   RESET_QUEUES, 0, 0);
+		}
 	}
 
 	if (adev->gfx.kiq[0].ring.sched.ready && !adev->job_hang)
@@ -549,22 +552,26 @@ int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id)
 	DRM_INFO("kiq ring mec %d pipe %d q %d\n", kiq_ring->me, kiq_ring->pipe,
 							kiq_ring->queue);
 	spin_lock(&kiq->ring_lock);
-	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
-					adev->gfx.num_compute_rings +
-					kiq->pmf->set_resources_size);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		spin_unlock(&adev->gfx.kiq[0].ring_lock);
-		return r;
-	}
+	/* No need to map kcq on the slave */
+	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+		r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
+						adev->gfx.num_compute_rings +
+						kiq->pmf->set_resources_size);
+		if (r) {
+			DRM_ERROR("Failed to lock KIQ (%d).\n", r);
+			spin_unlock(&adev->gfx.kiq[0].ring_lock);
+			return r;
+		}
 
-	if (adev->enable_mes)
-		queue_mask = ~0ULL;
+		if (adev->enable_mes)
+			queue_mask = ~0ULL;
 
-	kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
-	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		j = i + xcc_id * adev->gfx.num_compute_rings;
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.compute_ring[i]);
+		kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_compute_rings;
+			kiq->pmf->kiq_map_queues(kiq_ring,
+						 &adev->gfx.compute_ring[i]);
+		}
 	}
 
 	r = amdgpu_ring_test_helper(kiq_ring);
@@ -1078,3 +1085,9 @@ void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev,
 		adev->firmware.fw_size += ALIGN(fw_size, PAGE_SIZE);
 	}
 }
+
+bool amdgpu_gfx_is_master_xcc(struct amdgpu_device *adev, int xcc_id)
+{
+	return !(xcc_id % (adev->gfx.num_xcc_per_xcp ?
+			adev->gfx.num_xcc_per_xcp : 1));
+}