diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index 3632963ac..03bebfa07 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -600,6 +600,7 @@ static void gk20a_remove_fifo_support(struct fifo_gk20a *f)
 	}
 
 	gk20a_fifo_delete_runlist(f);
+	nvgpu_mutex_destroy(&f->runlist_submit_mutex);
 
 	nvgpu_kfree(g, f->pbdma_map);
 	f->pbdma_map = NULL;
@@ -919,6 +920,12 @@ int gk20a_init_fifo_setup_sw_common(struct gk20a *g)
 		return err;
 	}
 
+	err = nvgpu_mutex_init(&f->runlist_submit_mutex);
+	if (err) {
+		nvgpu_err(g, "failed to init runlist_submit_mutex");
+		return err;
+	}
+
 	g->ops.fifo.init_pbdma_intr_descs(f); /* just filling in data/tables */
 
 	f->num_channels = g->ops.fifo.get_num_fifos(g);
@@ -3583,6 +3590,7 @@ int gk20a_fifo_update_runlist_locked(struct gk20a *g, u32 runlist_id,
 		runlist->count = 0;
 	}
 
+	nvgpu_mutex_acquire(&f->runlist_submit_mutex);
 	g->ops.fifo.runlist_hw_submit(g, runlist_id, runlist->count, new_buf);
 
 	if (wait_for_finish) {
@@ -3590,6 +3598,7 @@ int gk20a_fifo_update_runlist_locked(struct gk20a *g, u32 runlist_id,
 
 		if (ret == -ETIMEDOUT) {
 			nvgpu_err(g, "runlist %d update timeout", runlist_id);
+			nvgpu_mutex_release(&f->runlist_submit_mutex);
 			/* trigger runlist update timeout recovery */
 			return ret;
 
@@ -3597,6 +3606,7 @@ int gk20a_fifo_update_runlist_locked(struct gk20a *g, u32 runlist_id,
 			nvgpu_err(g, "runlist update interrupted");
 		}
 	}
+	nvgpu_mutex_release(&f->runlist_submit_mutex);
 
 	runlist->cur_buffer = new_buf;
 
diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h
index 21922426c..60e8998a7 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h
@@ -185,6 +185,7 @@ struct fifo_gk20a {
 	struct nvgpu_list_node free_chs;
 	struct nvgpu_mutex free_chs_mutex;
 	struct nvgpu_mutex gr_reset_mutex;
+	struct nvgpu_mutex runlist_submit_mutex;
 
 	struct tsg_gk20a *tsg;
 	struct nvgpu_mutex tsg_inuse_mutex;