01:27 <stapnewb> What does "!" in the following context: probe scheduler.wakeup = kernel.trace("sched_wakeup") !, kernel.function("try_to_wake_up")

01:27 <stapnewb> This is from https://github.com/jav/systemtap/blob/master/tapset/scheduler.stp

01:34 <fche> hey

01:35 <fche> it means that the scheduler.wakeup probe alias prefers the former alternative (a kernel tracepoint) to the other option(s)

01:35 <fche> so if kernel.trace("sched_wakeup") exists, it will ignore the others

01:35 <stapnewb> fche: Thank you very much!

01:35 <fche> this is briefly discussed in [man stapprobes]

01:37 <fche> (it's a way of making a tapset adapt to a variety of kernel generations / configurations)

01:38 <stapnewb> fche: Got it! I was looking at the SystemTap language reference and couldn't find it in there. Thanks again!

01:39 <fche> hm it belongs in there too

05:33 <introom> fche: common to plug systmetap into a running kernel? will systemtap itself cause much overhead?

07:51 <introom> or to say, is it unacceptable to plug systemtap into a production server to modify some parts of kernel logic ?

09:59 <fche> introom, this sort of thing is done as necessary; the overhead depends entirely on the choice of script