[PATCH 02/32] drm/amdgpu: enable the ring and IB test for slave kcq

Alex Deucher alexander.deucher at amd.com
Tue Mar 28 15:13:14 UTC 2023


From: Shiwu Zhang <shiwu.zhang at amd.com>

With the mec FW update to utilize the mqd base set by
driver for kcq mapping, slave kcq ring test and IB test
can be re-enabled.

Signed-off-by: Shiwu Zhang <shiwu.zhang at amd.com>
Reviewed-by: Le Ma <Le.Ma at amd.com>
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 61 ++++++++++++-------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c  |  5 --
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 12 ++---
 3 files changed, 33 insertions(+), 45 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 14ea9bbc3715..5ef6fbe354c9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -442,7 +442,7 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 
 			/* prepare MQD backup */
 			adev->gfx.mec.mqd_backup[i + xcc_id * adev->gfx.num_compute_rings] = kmalloc(mqd_size, GFP_KERNEL);
-			if (!adev->gfx.mec.mqd_backup[i])
+			if (!adev->gfx.mec.mqd_backup[i + xcc_id * adev->gfx.num_compute_rings])
 				dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
 		}
 	}
@@ -468,8 +468,8 @@ void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev, int xcc_id)
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
 		j = i + xcc_id * adev->gfx.num_compute_rings;
-		ring = &adev->gfx.compute_ring[i];
-		kfree(adev->gfx.mec.mqd_backup[i]);
+		ring = &adev->gfx.compute_ring[j];
+		kfree(adev->gfx.mec.mqd_backup[j]);
 		amdgpu_bo_free_kernel(&ring->mqd_obj,
 				      &ring->mqd_gpu_addr,
 				      &ring->mqd_ptr);
@@ -494,22 +494,20 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
 		return -EINVAL;
 
 	spin_lock(&kiq->ring_lock);
-	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;
-		}
+	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)
+	if (kiq_ring->sched.ready && !adev->job_hang)
 		r = amdgpu_ring_test_helper(kiq_ring);
 	spin_unlock(&kiq->ring_lock);
 
@@ -557,26 +555,23 @@ 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);
-	/* 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;
-		}
+	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(&kiq->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_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]);
-		}
+						 &adev->gfx.compute_ring[j]);
 	}
 
 	r = amdgpu_ring_test_helper(kiq_ring);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index 15fd61c202c8..2b6fdcf3bf60 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -407,11 +407,6 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 		else
 			tmo = tmo_gfx;
 
-		/* skip ib test on the slave kcq */
-		if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE &&
-		    !amdgpu_gfx_is_master_xcc(adev, ring->xcc_id))
-			continue;
-
 		r = amdgpu_ring_test_ib(ring, tmo);
 		if (!r) {
 			DRM_DEV_DEBUG(adev->dev, "ib test on %s succeeded\n",
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index 351bc16b95ae..5cc4c2c31b3a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -1962,13 +1962,11 @@ static int gfx_v9_4_3_cp_resume(struct amdgpu_device *adev)
 		if (r)
 			return r;
 
-		/* skip ring test on slave kcq */
-		if (amdgpu_gfx_is_master_xcc(adev, i)) {
-			for (j = 0; j < adev->gfx.num_compute_rings; j++) {
-				ring = &adev->gfx.compute_ring[j +
-					i * adev->gfx.num_compute_rings];
-				amdgpu_ring_test_helper(ring);
-			}
+		for (j = 0; j < adev->gfx.num_compute_rings; j++) {
+			ring = &adev->gfx.compute_ring[j + i * adev->gfx.num_compute_rings];
+			r = amdgpu_ring_test_helper(ring);
+			if (r)
+				return r;
 		}
 
 		gfx_v9_4_3_enable_gui_idle_interrupt(adev, true, i);
-- 
2.39.2



More information about the amd-gfx mailing list