The discussion started out as a straightforward
patch set from Thomas Gleixner making a minor change to how preemption
counting is handled. The resulting discussion quickly spread out to cover
a number of issues relevant to core-kernel development in surprisingly few
messages; each of those topics merits a quick look, starting with how the
preemption counter itself works. Sometimes a simple count turns out to not
be as simple as it seems.

Source: LWN.net – [$] Four short stories about preempt_count()