14:48 <JohnDoe__> Hi, is there anyone that used raw performance counters in systemtap (like those specified via -rXXXXX to perf command)?

14:53 <fche> hi JohnDoe__

14:54 <fche> sort of -- [man stapprobes] shows how perf.type(NN).config(MM) numbers may be specified raw

14:54 <fche> but I don't think we expose all the possibilities from the underyling API

15:11 <JohnDoe__> I already went through the stapprobes docs, identified the "type" for RAW counters, but I'm not sure how should I pass the raw counter identifier (as generated by by libpfm)

15:12 <JohnDoe__> tried to convert it from hex to decimal and pass as "config"

15:12 <JohnDoe__> but the values I get make no sense...

15:12 <fche> the type + config numbers both would need to be set correctly

15:13 <fche> could try stracing perf to see how it sets up those perfevent_attr structs

15:15 <JohnDoe__> well.. the question is how? I was looking for a stap probe that would allow for attaching to perf_event_open to explore the struct but seems like there is no such probe point...

15:15 <fche> sys_perf_event_open methinks

15:15 <fche> or just /usr/sbin/strace perf ... for real

15:16 <JohnDoe__> is the perf_event_open an syscall?

15:20 <JohnDoe__> oh - it really is

15:20 <fche> yup

15:20 <JohnDoe__> now have to explore what it shows

15:20 <JohnDoe__> thanks for the tip!

15:21 <fche> righto

15:40 <irker240> systemtap: dsmith systemtap.git:refs/heads/master * release-3.1-222-g109c21b / httpd/api.cxx httpd/api.h httpd/backends.cxx httpd/client.py util.cxx util.h: Send over the linux distribution info to the http server. http://tinyurl.com/yc9854re

16:16 <irker240> systemtap: fche systemtap.git:refs/heads/master * release-3.1-223-g536ef4b / stapbpf/stapbpf.cxx: add PRERELEASE marker to stapbpf startup line http://tinyurl.com/yagspzqk