diff --git a/flang/lib/Optimizer/Transforms/CUFOpConversion.cpp b/flang/lib/Optimizer/Transforms/CUFOpConversion.cpp
index 9de20f0f0d45e..17699dadc7511 100644
--- a/flang/lib/Optimizer/Transforms/CUFOpConversion.cpp
+++ b/flang/lib/Optimizer/Transforms/CUFOpConversion.cpp
@@ -654,7 +654,7 @@ struct CUFDataTransferOpConversion
               loc, builder);
       }
       auto materializeBoxIfNeeded = [&](mlir::Value val) -> mlir::Value {
-        if (mlir::isa<fir::EmboxOp>(val.getDefiningOp())) {
+        if (mlir::isa<fir::EmboxOp, fir::ReboxOp>(val.getDefiningOp())) {
           // Materialize the box to memory to be able to call the runtime.
           mlir::Value box = builder.createTemporary(loc, val.getType());
           builder.create<fir::StoreOp>(loc, val, box);
diff --git a/flang/test/Fir/CUDA/cuda-data-transfer.fir b/flang/test/Fir/CUDA/cuda-data-transfer.fir
index 1ee44f3c6d97c..5f10dc0562d17 100644
--- a/flang/test/Fir/CUDA/cuda-data-transfer.fir
+++ b/flang/test/Fir/CUDA/cuda-data-transfer.fir
@@ -466,4 +466,51 @@ func.func @_QPlogical_cst() {
 // CHECK: %[[BOX_NONE:.*]] = fir.convert %[[DESC]] : (!fir.ref<!fir.box<!fir.logical<4>>>) -> !fir.ref<!fir.box<none>>
 // CHECK: fir.call @_FortranACUFDataTransferCstDesc(%{{.*}}, %[[BOX_NONE]], %{{.*}}, %{{.*}}, %{{.*}}) : (!fir.ref<!fir.box<none>>, !fir.ref<!fir.box<none>>, i32, !fir.ref<i8>, i32) -> none
 
+func.func @_QPcallkernel(%arg0: !fir.box<!fir.array<?x?xcomplex<f32>>> {fir.bindc_name = "a"}, %arg1: !fir.ref<f32> {fir.bindc_name = "b"}, %arg2: !fir.ref<f32> {fir.bindc_name = "c"}) {
+  %c0_i64 = arith.constant 0 : i64
+  %c1_i32 = arith.constant 1 : i32
+  %c0_i32 = arith.constant 0 : i32
+  %c1 = arith.constant 1 : index
+  %c0 = arith.constant 0 : index
+  %0 = fir.dummy_scope : !fir.dscope
+  %1 = fir.declare %arg0 dummy_scope %0 {uniq_name = "_QFcallkernelEa"} : (!fir.box<!fir.array<?x?xcomplex<f32>>>, !fir.dscope) -> !fir.box<!fir.array<?x?xcomplex<f32>>>
+  %2 = fir.rebox %1 : (!fir.box<!fir.array<?x?xcomplex<f32>>>) -> !fir.box<!fir.array<?x?xcomplex<f32>>>
+  %3 = cuf.alloc !fir.box<!fir.heap<!fir.array<?x?xcomplex<f32>>>> {bindc_name = "adev", data_attr = #cuf.cuda<device>, uniq_name = "_QFcallkernelEadev"} -> !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xcomplex<f32>>>>>
+  %7 = fir.declare %3 {data_attr = #cuf.cuda<device>, fortran_attrs = #fir.var_attrs<allocatable>, uniq_name = "_QFcallkernelEadev"} : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xcomplex<f32>>>>>) -> !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xcomplex<f32>>>>>
+  %8 = fir.declare %arg1 dummy_scope %0 {uniq_name = "_QFcallkernelEb"} : (!fir.ref<f32>, !fir.dscope) -> !fir.ref<f32>
+  %9 = fir.declare %arg2 dummy_scope %0 {uniq_name = "_QFcallkernelEc"} : (!fir.ref<f32>, !fir.dscope) -> !fir.ref<f32>
+  %10 = fir.alloca i32 {bindc_name = "m", uniq_name = "_QFcallkernelEm"}
+  %11 = fir.declare %10 {uniq_name = "_QFcallkernelEm"} : (!fir.ref<i32>) -> !fir.ref<i32>
+  %12 = fir.alloca i32 {bindc_name = "n", uniq_name = "_QFcallkernelEn"}
+  %13 = fir.declare %12 {uniq_name = "_QFcallkernelEn"} : (!fir.ref<i32>) -> !fir.ref<i32>
+  %14:3 = fir.box_dims %2, %c0 : (!fir.box<!fir.array<?x?xcomplex<f32>>>, index) -> (index, index, index)
+  %15 = fir.convert %14#1 : (index) -> i32
+  fir.store %15 to %13 : !fir.ref<i32>
+  %16:3 = fir.box_dims %2, %c1 : (!fir.box<!fir.array<?x?xcomplex<f32>>>, index) -> (index, index, index)
+  %27 = fir.load %13 : !fir.ref<i32>
+  %28 = fir.convert %27 : (i32) -> index
+  %29 = arith.cmpi sgt, %28, %c0 : index
+  %30 = arith.select %29, %28, %c0 : index
+  %31 = fir.load %11 : !fir.ref<i32>
+  %32 = fir.convert %31 : (i32) -> index
+  %33 = arith.cmpi sgt, %32, %c0 : index
+  %34 = arith.select %33, %32, %c0 : index
+  %35 = fir.shape %30, %34 : (index, index) -> !fir.shape<2>
+  %36 = fir.undefined index
+  %37 = fir.slice %c1, %28, %c1, %c1, %32, %c1 : (index, index, index, index, index, index) -> !fir.slice<2>
+  %38 = fir.rebox %2 [%37] : (!fir.box<!fir.array<?x?xcomplex<f32>>>, !fir.slice<2>) -> !fir.box<!fir.array<?x?xcomplex<f32>>>
+  cuf.data_transfer %38 to %7 {transfer_kind = #cuf.cuda_transfer<host_device>} : !fir.box<!fir.array<?x?xcomplex<f32>>>, !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xcomplex<f32>>>>>
+  return
+}
+
+// CHECK-LABEL: func.func @_QPcallkernel(
+// CHECK-SAME: %[[ARG0:.*]]: !fir.box<!fir.array<?x?xcomplex<f32>>> {fir.bindc_name = "a"}
+// CHECK: %[[ALLOCA:.*]] = fir.alloca !fir.box<!fir.array<?x?xcomplex<f32>>>
+// CHECK: %[[DECL_ARG0:.*]] = fir.declare %[[ARG0]] dummy_scope %{{.*}} {uniq_name = "_QFcallkernelEa"} : (!fir.box<!fir.array<?x?xcomplex<f32>>>, !fir.dscope) -> !fir.box<!fir.array<?x?xcomplex<f32>>>
+// CHECK: %[[REBOX0:.*]] = fir.rebox %[[DECL_ARG0]] : (!fir.box<!fir.array<?x?xcomplex<f32>>>) -> !fir.box<!fir.array<?x?xcomplex<f32>>>
+// CHECK: %[[REBOX1:.*]] = fir.rebox %[[REBOX0]] [%{{.*}}] : (!fir.box<!fir.array<?x?xcomplex<f32>>>, !fir.slice<2>) -> !fir.box<!fir.array<?x?xcomplex<f32>>>
+// CHECK: fir.store %[[REBOX1]] to %[[ALLOCA]] : !fir.ref<!fir.box<!fir.array<?x?xcomplex<f32>>>>
+// CHECK: %[[BOX_NONE:.*]] = fir.convert %[[ALLOCA]] : (!fir.ref<!fir.box<!fir.array<?x?xcomplex<f32>>>>) -> !fir.ref<!fir.box<none>>
+// CHECK: fir.call @_FortranACUFDataTransferDescDesc(%{{.*}}, %[[BOX_NONE]], %{{.*}}, %{{.*}}, %{{.*}}) : (!fir.ref<!fir.box<none>>, !fir.ref<!fir.box<none>>, i32, !fir.ref<i8>, i32) -> none
+
 } // end of module