On Thu, Sep 5, 2019 at 3:12 PM Daniel Colascione <dancol@xxxxxxxxxx> wrote: > Basically, what I have in mind is this: Actually --- I wonder whether there's already enough power in the trigger mechanism to do this without any code changes to ftrace histograms themselves. I'm trying to think of the minimum additional kernel facility that we'd need to implement the scheme I described above, and it might be that we don't need to do anything at all except add the actual level tracepoints.