On Wed, Feb 12, 2020 at 10:14:15AM -0500, Steven Rostedt wrote:

quoted

My initial patch has __preempt_count_add/sub() in, but then I figured
someone would go complain the tracepoint would go missing.

Fine, but what bug are you trying to fix? I haven't seen one mentioned
yet. Function tracing has recursion protection, and tracing
preempt_count in nmi_enter() causes no problems. What's the problem you
are trying to solve?

The same one as yesterday; if we hit a tracer in NMI context, when
!rcu_is_watching() and in_nmi() is still 0, we're fucked.