I agree with Erik: generaly, watchdog must not be disabled and it must not be a way to disable it anyhow.
There is special ideology in high-stable system based on non-disabled watchdog. The main idea is that each part of a program executes in watchdog timeout and then stops till watchdog reset. After each reset a task dispatcher checks for that previous task has been done successful and if so then runs next program part. It is really stable sequence because it utilizes reset. It may be used on 51`s - thanks that RAM-content is not affected with reset. But we need to have non-disabled watchdog (and I preffer to have watchdog which is active after any reset itself).
Good days!