diff --git a/drivers/gpu/nvgpu/common/init/nvgpu_init.c b/drivers/gpu/nvgpu/common/init/nvgpu_init.c
index 7ca1c48fb..92603649f 100644
--- a/drivers/gpu/nvgpu/common/init/nvgpu_init.c
+++ b/drivers/gpu/nvgpu/common/init/nvgpu_init.c
@@ -339,10 +339,6 @@ int nvgpu_prepare_poweroff(struct gk20a *g)
 	if (tmp_ret != 0) {
 		ret = tmp_ret;
 	}
-	tmp_ret = nvgpu_nvs_suspend(g);
-	if (tmp_ret != 0) {
-		ret = tmp_ret;
-	}
 	tmp_ret = g->ops.fifo.fifo_suspend(g);
 	if (tmp_ret != 0) {
 		ret = tmp_ret;
diff --git a/drivers/gpu/nvgpu/common/nvs/nvs_sched.c b/drivers/gpu/nvgpu/common/nvs/nvs_sched.c
index f393f19aa..b456b73a8 100644
--- a/drivers/gpu/nvgpu/common/nvs/nvs_sched.c
+++ b/drivers/gpu/nvgpu/common/nvs/nvs_sched.c
@@ -161,6 +161,8 @@ static void nvgpu_nvs_worker_deinit(struct gk20a *g)
 	struct nvgpu_worker *worker = &g->scheduler->worker.worker;
 
 	nvgpu_worker_deinit(worker);
+
+	nvs_dbg(g, "NVS worker suspended");
 }
 
 int nvgpu_nvs_init(struct gk20a *g)
@@ -198,6 +200,8 @@ void nvgpu_nvs_remove_support(struct gk20a *g)
 		return;
 	}
 
+	nvgpu_nvs_worker_deinit(g);
+
 	nvs_domain_for_each(sched->sched, nvs_dom) {
 		struct nvgpu_nvs_domain *nvgpu_dom = nvs_dom->priv;
 		if (nvgpu_dom->ref != 1U) {
@@ -217,14 +221,6 @@ void nvgpu_nvs_remove_support(struct gk20a *g)
 	nvgpu_mutex_destroy(&g->sched_mutex);
 }
 
-int nvgpu_nvs_suspend(struct gk20a *g)
-{
-	nvgpu_nvs_worker_deinit(g);
-	nvs_dbg(g, "NVS worker suspended");
-
-	return 0;
-}
-
 int nvgpu_nvs_open(struct gk20a *g)
 {
 	int err = 0;
@@ -235,8 +231,6 @@ int nvgpu_nvs_open(struct gk20a *g)
 
 	if (g->scheduler != NULL) {
 		/* resuming from railgate */
-		err = nvgpu_nvs_worker_init(g);
-		nvs_dbg(g, "NVS worker resume, err=%d", err);
 		goto unlock;
 	}
 
diff --git a/drivers/gpu/nvgpu/common/utils/worker.c b/drivers/gpu/nvgpu/common/utils/worker.c
index d5498cc78..a5bbe655e 100644
--- a/drivers/gpu/nvgpu/common/utils/worker.c
+++ b/drivers/gpu/nvgpu/common/utils/worker.c
@@ -199,6 +199,11 @@ static int nvgpu_worker_start(struct nvgpu_worker *worker)
 
 	err = nvgpu_thread_create(&worker->poll_task, worker,
 			nvgpu_worker_poll_work, worker->thread_name);
+	if (err != 0) {
+		nvgpu_err(worker->g,
+			  "failed to create worker poller thread %s err %d",
+			  worker->thread_name, err);
+	}
 
 	nvgpu_mutex_release(&worker->start_lock);
 	return err;
diff --git a/drivers/gpu/nvgpu/include/nvgpu/nvs.h b/drivers/gpu/nvgpu/include/nvgpu/nvs.h
index df51d46c3..ff112f0dc 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/nvs.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/nvs.h
@@ -84,7 +84,6 @@ struct nvgpu_nvs_scheduler {
 int nvgpu_nvs_init(struct gk20a *g);
 int nvgpu_nvs_open(struct gk20a *g);
 void nvgpu_nvs_remove_support(struct gk20a *g);
-int nvgpu_nvs_suspend(struct gk20a *g);
 void nvgpu_nvs_get_log(struct gk20a *g, s64 *timestamp, const char **msg);
 u32 nvgpu_nvs_domain_count(struct gk20a *g);
 int nvgpu_nvs_del_domain(struct gk20a *g, u64 dom_id);
@@ -113,11 +112,6 @@ static inline void nvgpu_nvs_remove_support(struct gk20a *g)
 {
 }
 
-static inline int nvgpu_nvs_suspend(struct gk20a *g)
-{
-	return 0;
-}
-
 static inline struct nvgpu_nvs_domain *
 nvgpu_nvs_domain_by_name(struct gk20a *g, const char *name)
 {