Aubrey Li wrote: > 2008/5/19 Rafael Vanoni <Rafael.Vanoni at sun.com>: >> I've added xcalls to the event report and the number of wakeups x events are >> more cohesive. This diff takes a shot at reporting that, let me know what >> you think. >> >> A related problem is the number of xcalls that libdtrace is generating, >> which puts powertop higher in the xcalls report than we'd like. I've been >> looking at this and emailed dtrace-discuss about it. >> > Just said on dtrace-discuss, you can use the dtrace stack() to find > the root cause. > > Yes, xcall is an event powertop should report, currently xcall dtrace probe > doesn't take the function as an argument, that prevents us getting the > caller easily. > Is it possible to change the current probe?
That would have to be discussed with the dtrace team. We could associate syscalls with xcalls, giving a more in depth feedback than just execname. But I wonder if the target user would prefer the execname. > execname may be acceptable, but we should exclude execname=powertop, > what do you think? I agree. thanks Rafael
