nvme_change_ctrl_state() bug?

[Steve Wise]

Shouldn't the assigning of the state be inside the lock?  

.
        spin_unlock_irq(&ctrl->lock);

        if (changed)
                ctrl->state = new_state;

        return changed;