drm/i915/gvt: do not deliver a workload if its creation fails
authorYan Zhao <yan.y.zhao@intel.com>
Wed, 27 Mar 2019 04:54:51 +0000 (00:54 -0400)
committerZhenyu Wang <zhenyuw@linux.intel.com>
Fri, 29 Mar 2019 02:26:53 +0000 (10:26 +0800)
in workload creation routine, if any failure occurs, do not queue this
workload for delivery. if this failure is fatal, enter into failsafe
mode.

Fixes: 6d76303553ba ("drm/i915/gvt: Move common vGPU workload creation into scheduler.c")
Cc: stable@vger.kernel.org #4.19+
Cc: zhenyuw@linux.intel.com
Signed-off-by: Yan Zhao <yan.y.zhao@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
drivers/gpu/drm/i915/gvt/scheduler.c

index 159192c097cc7eb7424070e8cec052f3f5e5b1f7..05b953793316b28ac1fb19c902474e468ba828b0 100644 (file)
@@ -1486,8 +1486,9 @@ intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id,
                intel_runtime_pm_put_unchecked(dev_priv);
        }
 
-       if (ret && (vgpu_is_vm_unhealthy(ret))) {
-               enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
+       if (ret) {
+               if (vgpu_is_vm_unhealthy(ret))
+                       enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
                intel_vgpu_destroy_workload(workload);
                return ERR_PTR(ret);
        }