On Wed, Jun 13, 2012 at 09:59:27AM +0800, Axel Lin wrote: > Signed-off-by: Axel Lin No, this (and generally a large proportion of possible usages of devm_request_threaded_irq()) is a bad idea - we need to ensure that the interrupt cannot fire when we're not in a position to handle it which in thi case means after the regulator is unregistered. Since we're not doing anything to explicitly quiesce the interrupt other than free it we need to ensure that the free_irq() happens strictly before the regulator is unregistered.