in workload_thread, it should grab runtime pm wakelock and later
uncore forcewake get will check rpm wakelock held successfully.
otherwise, sometimes, rpm wakelock not hold and print call trace below:
 Call Trace:
  intel_uncore_forcewake_get+0x15/0x20 [i915]
  workload_thread+0x5f9/0x16f0 [i915]
  ? __switch_to_asm+0x34/0x70
  ? __switch_to_asm+0x40/0x70
  ? __switch_to_asm+0x34/0x70
  ? __switch_to_asm+0x40/0x70
  ? __switch_to_asm+0x34/0x70
  ? __switch_to+0x85/0x3f0
  ? __switch_to_asm+0x40/0x70
  ? do_wait_intr_irq+0x90/0x90
  kthread+0x121/0x140
  ? intel_vgpu_clean_workloads+0x100/0x100 [i915]
  ? kthread_park+0x90/0x90
  ret_from_fork+0x35/0x40
 --[ end trace 
86525f742a02e12c ]--
v2: adapted to use rpm structure.
Fixes: 251d46b0875c ("drm/i915/gvt: Pin the per-engine GVT shadow contexts")
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Xiaolin Zhang <xiaolin.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
 
        int ret;
        bool need_force_wake = (INTEL_GEN(gvt->dev_priv) >= 9);
        DEFINE_WAIT_FUNC(wait, woken_wake_function);
+       struct intel_runtime_pm *rpm = &gvt->dev_priv->runtime_pm;
 
        kfree(p);
 
                                workload->ring_id, workload,
                                workload->vgpu->id);
 
+               intel_runtime_pm_get(rpm);
+
                gvt_dbg_sched("ring id %d will dispatch workload %p\n",
                                workload->ring_id, workload);
 
                        intel_uncore_forcewake_put(&gvt->dev_priv->uncore,
                                        FORCEWAKE_ALL);
 
+               intel_runtime_pm_put_unchecked(rpm);
                if (ret && (vgpu_is_vm_unhealthy(ret)))
                        enter_failsafe_mode(vgpu, GVT_FAILSAFE_GUEST_ERR);
        }