A couple of spinlock facts

Uniprocessor kernels optimize out spinlocks
Locking is a no-op

Holding a spinlock disables preemption
Required for deadlock avoidance

Long lock hold times can create latency problems
Even for unrelated code