Commit fdbd6948 authored by Jiadong Zhu's avatar Jiadong Zhu Committed by Alex Deucher
Browse files

drm/amdgpu/gfx9: wait for reset done before remap



There is a racing condition that cp firmware modifies
MQD in reset sequence after driver updates it for
remapping. We have to wait till CP_HQD_ACTIVE becoming
false then remap the queue.

v2: fix KIQ locking (Alex)
v3: fix KIQ locking harder

Acked-by: default avatarVitaly Prosyak <vitaly.prosyak@amd.com>
Signed-off-by: default avatarJiadong Zhu <Jiadong.Zhu@amd.com>
Reviewed-by: default avatarAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent b5e1a387
Loading
Loading
Loading
Loading
+31 −5
Original line number Diff line number Diff line
@@ -7125,7 +7125,7 @@ static int gfx_v9_0_reset_kcq(struct amdgpu_ring *ring,
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
	struct amdgpu_ring *kiq_ring = &kiq->ring;
	unsigned long flags;
	int r;
	int i, r;

	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
		return -EINVAL;
@@ -7147,9 +7147,28 @@ static int gfx_v9_0_reset_kcq(struct amdgpu_ring *ring,
	if (r)
		return r;

	/* make sure dequeue is complete*/
	gfx_v9_0_set_safe_mode(adev, 0);
	mutex_lock(&adev->srbm_mutex);
	soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
	for (i = 0; i < adev->usec_timeout; i++) {
		if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
			break;
		udelay(1);
	}
	if (i >= adev->usec_timeout)
		r = -ETIMEDOUT;
	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
	mutex_unlock(&adev->srbm_mutex);
	gfx_v9_0_unset_safe_mode(adev, 0);
	if (r) {
		dev_err(adev->dev, "fail to wait on hqd deactive\n");
		return r;
	}

	r = amdgpu_bo_reserve(ring->mqd_obj, false);
	if (unlikely(r != 0)){
		DRM_ERROR("fail to resv mqd_obj\n");
		dev_err(adev->dev, "fail to resv mqd_obj\n");
		return r;
	}
	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
@@ -7160,11 +7179,18 @@ static int gfx_v9_0_reset_kcq(struct amdgpu_ring *ring,
	}
	amdgpu_bo_unreserve(ring->mqd_obj);
	if (r) {
		DRM_ERROR("fail to unresv mqd_obj\n");
		dev_err(adev->dev, "fail to unresv mqd_obj\n");
		return r;
	}
	spin_lock_irqsave(&kiq->ring_lock, flags);
	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
	if (r) {
		spin_unlock_irqrestore(&kiq->ring_lock, flags);
		return -ENOMEM;
	}
	kiq->pmf->kiq_map_queues(kiq_ring, ring);
	amdgpu_ring_commit(kiq_ring);
	spin_unlock_irqrestore(&kiq->ring_lock, flags);
	r = amdgpu_ring_test_ring(kiq_ring);
	if (r) {
		DRM_ERROR("fail to remap queue\n");