drm/amdgpu: fix concurrent VM flushes on Vega/Navi v2

Starting with Vega the hardware supports concurrent flushes
of VMID which can be used to implement per process VMID
allocation.

But concurrent flushes are mutual exclusive with back to
back VMID allocations, fix this to avoid a VMID used in
two ways at the same time.

v2: don't set ring to NULL

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
Tested-by: James Zhu <James.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

3 files changed, 18 insertions, 8 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
index 94b069630db3..b4971e90b98c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
@@ -215,7 +215,11 @@ static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm,
 	/* Check if we have an idle VMID */
 	i = 0;
 	list_for_each_entry((*idle), &id_mgr->ids_lru, list) {
-		fences[i] = amdgpu_sync_peek_fence(&(*idle)->active, ring);
+		/* Don't use per engine and per process VMID at the same time */
+		struct amdgpu_ring *r = adev->vm_manager.concurrent_flush ?
+			NULL : ring;
+
+		fences[i] = amdgpu_sync_peek_fence(&(*idle)->active, r);
 		if (!fences[i])
 			break;
 		++i;
@@ -281,7 +285,7 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
 	if (updates && (*id)->flushed_updates &&
 	    updates->context == (*id)->flushed_updates->context &&
 	    !dma_fence_is_later(updates, (*id)->flushed_updates))
-	    updates = NULL;
+		updates = NULL;
 
 	if ((*id)->owner != vm->immediate.fence_context ||
 	    job->vm_pd_addr != (*id)->pd_gpu_addr ||
@@ -290,6 +294,10 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
 	     !dma_fence_is_signaled((*id)->last_flush))) {
 		struct dma_fence *tmp;
 
+		/* Don't use per engine and per process VMID at the same time */
+		if (adev->vm_manager.concurrent_flush)
+			ring = NULL;
+
 		/* to prevent one context starved by another context */
 		(*id)->pd_gpu_addr = 0;
 		tmp = amdgpu_sync_peek_fence(&(*id)->active, ring);
@@ -365,12 +373,7 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
 		if (updates && (!flushed || dma_fence_is_later(updates, flushed)))
 			needs_flush = true;
 
-		/* Concurrent flushes are only possible starting with Vega10 and
-		 * are broken on Navi10 and Navi14.
-		 */
-		if (needs_flush && (adev->asic_type < CHIP_VEGA10 ||
-				    adev->asic_type == CHIP_NAVI10 ||
-				    adev->asic_type == CHIP_NAVI14))
+		if (needs_flush && !adev->vm_manager.concurrent_flush)
 			continue;
 
 		/* Good, we can use this VMID. Remember this submission as
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 0ffdf847cad0..9acee4a5b2ba 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -3148,6 +3148,12 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 {
 	unsigned i;
 
+	/* Concurrent flushes are only possible starting with Vega10 and
+	 * are broken on Navi10 and Navi14.
+	 */
+	adev->vm_manager.concurrent_flush = !(adev->asic_type < CHIP_VEGA10 ||
+					      adev->asic_type == CHIP_NAVI10 ||
+					      adev->asic_type == CHIP_NAVI14);
 	amdgpu_vmid_mgr_init(adev);
 
 	adev->vm_manager.fence_context =
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 976a12e5a8b9..4e140288159c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -331,6 +331,7 @@ struct amdgpu_vm_manager {
 	/* Handling of VMIDs */
 	struct amdgpu_vmid_mgr			id_mgr[AMDGPU_MAX_VMHUBS];
 	unsigned int				first_kfd_vmid;
+	bool					concurrent_flush;
 
 	/* Handling of VM fences */
 	u64					fence_context;