diff --git a/drivers/gpu/nvgpu/common/falcon/falcon.c b/drivers/gpu/nvgpu/common/falcon/falcon.c
index 59a1f3daa..681ff105e 100644
--- a/drivers/gpu/nvgpu/common/falcon/falcon.c
+++ b/drivers/gpu/nvgpu/common/falcon/falcon.c
@@ -280,7 +280,9 @@ int nvgpu_falcon_copy_from_dmem(struct nvgpu_falcon *flcn,
 	flcn_ops = &flcn->flcn_ops;
 
 	if (flcn_ops->copy_from_dmem != NULL) {
+		nvgpu_mutex_acquire(&flcn->copy_lock);
 		status = flcn_ops->copy_from_dmem(flcn, src, dst, size, port);
+		nvgpu_mutex_release(&flcn->copy_lock);
 	} else {
 		nvgpu_warn(flcn->g, "Invalid op on falcon 0x%x ",
 			flcn->flcn_id);
@@ -302,7 +304,9 @@ int nvgpu_falcon_copy_to_dmem(struct nvgpu_falcon *flcn,
 	flcn_ops = &flcn->flcn_ops;
 
 	if (flcn_ops->copy_to_dmem != NULL) {
+		nvgpu_mutex_acquire(&flcn->copy_lock);
 		status = flcn_ops->copy_to_dmem(flcn, dst, src, size, port);
+		nvgpu_mutex_release(&flcn->copy_lock);
 	} else {
 		nvgpu_warn(flcn->g, "Invalid op on falcon 0x%x ",
 			flcn->flcn_id);
@@ -324,7 +328,9 @@ int nvgpu_falcon_copy_from_imem(struct nvgpu_falcon *flcn,
 	flcn_ops = &flcn->flcn_ops;
 
 	if (flcn_ops->copy_from_imem != NULL) {
+		nvgpu_mutex_acquire(&flcn->copy_lock);
 		status = flcn_ops->copy_from_imem(flcn, src, dst, size, port);
+		nvgpu_mutex_release(&flcn->copy_lock);
 	} else {
 		nvgpu_warn(flcn->g, "Invalid op on falcon 0x%x ",
 			flcn->flcn_id);
@@ -346,8 +352,10 @@ int nvgpu_falcon_copy_to_imem(struct nvgpu_falcon *flcn,
 	flcn_ops = &flcn->flcn_ops;
 
 	if (flcn_ops->copy_to_imem != NULL) {
+		nvgpu_mutex_acquire(&flcn->copy_lock);
 		status = flcn_ops->copy_to_imem(flcn, dst, src, size, port,
 					sec, tag);
+		nvgpu_mutex_release(&flcn->copy_lock);
 	} else {
 		nvgpu_warn(flcn->g, "Invalid op on falcon 0x%x ",
 			flcn->flcn_id);
@@ -609,6 +617,7 @@ int nvgpu_falcon_sw_init(struct gk20a *g, u32 flcn_id)
 {
 	struct nvgpu_falcon **flcn_p = NULL, *flcn = NULL;
 	struct gpu_ops *gops = &g->ops;
+	int err;
 
 	flcn_p = falcon_get_instance(g, flcn_id);
 	if (flcn_p == NULL) {
@@ -622,6 +631,13 @@ int nvgpu_falcon_sw_init(struct gk20a *g, u32 flcn_id)
 		return -ENOMEM;
 	}
 
+	err = nvgpu_mutex_init(&flcn->copy_lock);
+	if (err != 0) {
+		nvgpu_err(g, "Error in flcn.copy_lock mutex initialization");
+		nvgpu_kfree(g, flcn);
+		return err;
+	}
+
 	flcn->flcn_id = flcn_id;
 	flcn->g = g;
 
@@ -633,7 +649,7 @@ int nvgpu_falcon_sw_init(struct gk20a *g, u32 flcn_id)
 
 void nvgpu_falcon_sw_free(struct gk20a *g, u32 flcn_id)
 {
-	struct nvgpu_falcon **flcn_p = NULL;
+	struct nvgpu_falcon **flcn_p = NULL, *flcn = NULL;
 	struct gpu_ops *gops = &g->ops;
 
 	flcn_p = falcon_get_instance(g, flcn_id);
@@ -641,7 +657,9 @@ void nvgpu_falcon_sw_free(struct gk20a *g, u32 flcn_id)
 		return;
 	}
 
-	gops->falcon.falcon_hal_sw_free(*flcn_p);
-	nvgpu_kfree(g, *flcn_p);
+	flcn = *flcn_p;
+	gops->falcon.falcon_hal_sw_free(flcn);
+	nvgpu_mutex_destroy(&flcn->copy_lock);
+	nvgpu_kfree(g, flcn);
 	*flcn_p = NULL;
 }
diff --git a/drivers/gpu/nvgpu/common/falcon/falcon_gk20a.c b/drivers/gpu/nvgpu/common/falcon/falcon_gk20a.c
index 08d4829f5..6dd56fb2c 100644
--- a/drivers/gpu/nvgpu/common/falcon/falcon_gk20a.c
+++ b/drivers/gpu/nvgpu/common/falcon/falcon_gk20a.c
@@ -207,8 +207,6 @@ static int gk20a_falcon_copy_from_dmem(struct nvgpu_falcon *flcn,
 		return -EINVAL;
 	}
 
-	nvgpu_mutex_acquire(&flcn->copy_lock);
-
 	words = size >> 2;
 	bytes = size & 0x3U;
 
@@ -232,7 +230,6 @@ static int gk20a_falcon_copy_from_dmem(struct nvgpu_falcon *flcn,
 		}
 	}
 
-	nvgpu_mutex_release(&flcn->copy_lock);
 	return 0;
 }
 
@@ -252,8 +249,6 @@ static int gk20a_falcon_copy_to_dmem(struct nvgpu_falcon *flcn,
 		return -EINVAL;
 	}
 
-	nvgpu_mutex_acquire(&flcn->copy_lock);
-
 	words = size >> 2;
 	bytes = size & 0x3U;
 
@@ -286,8 +281,6 @@ static int gk20a_falcon_copy_to_dmem(struct nvgpu_falcon *flcn,
 			data - dst, size);
 	}
 
-	nvgpu_mutex_release(&flcn->copy_lock);
-
 	return 0;
 }
 
@@ -310,8 +303,6 @@ static int gk20a_falcon_copy_from_imem(struct nvgpu_falcon *flcn, u32 src,
 		return -EINVAL;
 	}
 
-	nvgpu_mutex_acquire(&flcn->copy_lock);
-
 	words = size >> 2;
 	bytes = size & 0x3U;
 	blk = src >> 8;
@@ -336,8 +327,6 @@ static int gk20a_falcon_copy_from_imem(struct nvgpu_falcon *flcn, u32 src,
 		}
 	}
 
-	nvgpu_mutex_release(&flcn->copy_lock);
-
 	return 0;
 }
 
@@ -358,8 +347,6 @@ static int gk20a_falcon_copy_to_imem(struct nvgpu_falcon *flcn, u32 dst,
 		return -EINVAL;
 	}
 
-	nvgpu_mutex_acquire(&flcn->copy_lock);
-
 	words = size >> 2;
 	blk = dst >> 8;
 
@@ -391,8 +378,6 @@ static int gk20a_falcon_copy_to_imem(struct nvgpu_falcon *flcn, u32 dst,
 		i++;
 	}
 
-	nvgpu_mutex_release(&flcn->copy_lock);
-
 	return 0;
 }
 
@@ -753,12 +738,7 @@ int gk20a_falcon_hal_sw_init(struct nvgpu_falcon *flcn)
 	}
 
 	if (flcn->is_falcon_supported) {
-		err = nvgpu_mutex_init(&flcn->copy_lock);
-		if (err != 0) {
-			nvgpu_err(g, "Error in flcn.copy_lock mutex initialization");
-		} else {
-			gk20a_falcon_ops(flcn);
-		}
+		gk20a_falcon_ops(flcn);
 	} else {
 		nvgpu_log_info(g, "falcon 0x%x not supported on %s",
 			flcn->flcn_id, g->name);
@@ -772,7 +752,6 @@ void gk20a_falcon_hal_sw_free(struct nvgpu_falcon *flcn)
 	struct gk20a *g = flcn->g;
 
 	if (flcn->is_falcon_supported) {
-		nvgpu_mutex_destroy(&flcn->copy_lock);
 		flcn->is_falcon_supported = false;
 	} else {
 		nvgpu_log_info(g, "falcon 0x%x not supported on %s",
diff --git a/drivers/gpu/nvgpu/common/falcon/falcon_gp106.c b/drivers/gpu/nvgpu/common/falcon/falcon_gp106.c
index 321028db0..deaa5ac7a 100644
--- a/drivers/gpu/nvgpu/common/falcon/falcon_gp106.c
+++ b/drivers/gpu/nvgpu/common/falcon/falcon_gp106.c
@@ -90,12 +90,7 @@ int gp106_falcon_hal_sw_init(struct nvgpu_falcon *flcn)
 	}
 
 	if (flcn->is_falcon_supported) {
-		err = nvgpu_mutex_init(&flcn->copy_lock);
-		if (err != 0) {
-			nvgpu_err(g, "Error in copy_lock mutex initialization");
-		} else {
-			gp106_falcon_ops(flcn);
-		}
+		gp106_falcon_ops(flcn);
 	} else {
 		nvgpu_info(g, "falcon 0x%x not supported on %s",
 			flcn->flcn_id, g->name);
diff --git a/drivers/gpu/nvgpu/common/falcon/falcon_gv100.c b/drivers/gpu/nvgpu/common/falcon/falcon_gv100.c
index e907ca0f5..bfc31ec02 100644
--- a/drivers/gpu/nvgpu/common/falcon/falcon_gv100.c
+++ b/drivers/gpu/nvgpu/common/falcon/falcon_gv100.c
@@ -70,12 +70,7 @@ int gv100_falcon_hal_sw_init(struct nvgpu_falcon *flcn)
 	}
 
 	if (flcn->is_falcon_supported) {
-		err = nvgpu_mutex_init(&flcn->copy_lock);
-		if (err != 0) {
-			nvgpu_err(g, "Error in flcn.copy_lock mutex initialization");
-		} else {
-			gv100_falcon_ops(flcn);
-		}
+		gv100_falcon_ops(flcn);
 	} else {
 		/*
 		 * Forward call to previous chips HAL
diff --git a/drivers/gpu/nvgpu/common/falcon/falcon_tu104.c b/drivers/gpu/nvgpu/common/falcon/falcon_tu104.c
index 5be1c69a6..fb55ca45e 100644
--- a/drivers/gpu/nvgpu/common/falcon/falcon_tu104.c
+++ b/drivers/gpu/nvgpu/common/falcon/falcon_tu104.c
@@ -79,12 +79,7 @@ int tu104_falcon_hal_sw_init(struct nvgpu_falcon *flcn)
 	}
 
 	if (flcn->is_falcon_supported) {
-		err = nvgpu_mutex_init(&flcn->copy_lock);
-		if (err != 0) {
-			nvgpu_err(g, "Error in flcn.copy_lock mutex initialization");
-		} else {
-			tu104_falcon_ops(flcn);
-		}
+		tu104_falcon_ops(flcn);
 	} else {
 		/*
 		 * Forward call to previous chips HAL