Currently, the RSA implementation requires a large stack size. Since RSA computation is likely to take place only once at a certain time (e.g. boot time), it's better to use a shared memory as a work buffer than allocating a large stack to a task.