diff --git a/ports/cortex-m/atomport.c b/ports/cortex-m/atomport.c
index ba4a4bf..49ea73d 100644
--- a/ports/cortex-m/atomport.c
+++ b/ports/cortex-m/atomport.c
@@ -143,6 +143,11 @@ void archThreadContextInit(ATOM_TCB *tcb_ptr, void *stack_top,
     struct isr_stack *isr_ctx;
     struct task_stack *tsk_ctx;
 
+    /**
+     * Enforce initial stack alignment
+     */
+    stack_top = STACK_ALIGN(stack_top, STACK_ALIGN_SIZE);
+
     /**
      * New threads will be scheduled from an exception handler, so we have to
      * set up an exception stack frame as well as task stack frame
diff --git a/ports/cortex-m/atomport.h b/ports/cortex-m/atomport.h
index 1aba5c0..562b84e 100644
--- a/ports/cortex-m/atomport.h
+++ b/ports/cortex-m/atomport.h
@@ -40,6 +40,10 @@
 /* Size of each stack entry / stack alignment size (4 bytes on Cortex-M without FPU) */
 #define STACK_ALIGN_SIZE    sizeof(uint32_t)
 
+#define ALIGN(x, a)         ((x + (typeof(x))(a) - 1) & ~((typeof(x))(a) - 1))
+#define PTR_ALIGN(p, a)     ((typeof(p))ALIGN((uint32_t)(p), (a)))
+#define STACK_ALIGN(p, a)   (typeof(p))((typeof(a))(p) & ~((a) - 1))
+
 #define POINTER             void *
 #define UINT32              uint32_t