14:47 <irker275> systemtap: dsmith systemtap.git:refs/heads/master * release-3.1-313-g2cd9b44 / testsuite/systemtap.base/stap_hwbkpt_kmod.c testsuite/systemtap.base/stap_kmodule.c testsuite/systemtap.context/systemtap_test_module1.c testsuite/systemtap.printf/memory1_module.c: Add some testsuite test kernel module fixes for rawhide. http://tinyurl.com/yb38rght

15:20 <irker275> systemtap: dsmith systemtap.git:refs/heads/master * release-3.1-314-g11784d6 / tapset/linux/sysc_msgctl.stp tapset/linux/sysc_shmctl.stp: Updated the sysc_msgctl and sysc_shmctl tapsets for rawhide. http://tinyurl.com/ycgf4hhu

17:45 <irker275> systemtap: amerey systemtap.git:refs/heads/master * release-3.1-315-ge2298f9 / testsuite/systemtap.bpf/bpf_tests/array.stp testsuite/systemtap.bpf/bpf_tests/while.stp: stapbpf: add while loop test http://tinyurl.com/yca9ycr8

17:45 <irker275> systemtap: amerey systemtap.git:refs/heads/master * release-3.1-316-g591bc40 / NEWS doc/Makefile.in java/Makefile.in man/stap.1.in stapbpf/Makefile.am stapbpf/Makefile.in: Add stapbpf man page, update NEWS and stap man page with stapbpf info. http://tinyurl.com/y8bwmh8d

17:51 <irker275> systemtap: amerey systemtap.git:refs/heads/master * release-3.1-317-gbbd8aae / stapbpf/stapbpf.8: Add stapbpf.8 to stapbpf directory. http://tinyurl.com/ycymeg59

19:13 <silvaleandrodp> Hi everyone

19:14 <silvaleandrodp> i'm having a hardtime with probe perf someone could help?

19:29 <fche> silvaleandrodp, can try, what's up?

19:33 <silvaleandrodp> Hi fche i'm trying to get the return of a perf counter but it seems that perf.type(0).config(0) don't have any variables

19:33 <fche> a probe perf.* doesn't let you -read- the perfcounter value; it just arranges to interrupt / run a handler every XXX counts

19:34 <fche> that's because, at least at the time this feature was written, the kernel did not possess a suitable api for us to read actual counter values from our execution contexts

19:34 <fche> there is another similar facility, for userspace probes only, that lets one read process-specific perfcounters

19:35 <fche> see [man stapprobes], the probe perf.type(NN).config(MM).process("PROC").counter("COUNTER") { } part and process("PROCESS").statement("func@file") {stat <<< @perf("NAME")}

19:36 <irker275> systemtap: fche systemtap.git:refs/heads/master * release-3.1-318-g3c612a1 / man/stapprobes.3stap: man stapprobes: tweak uprobe-perf probe templates to match metavariables http://tinyurl.com/ybgte8je

19:40 <silvaleandrodp> I see, I will try these, i'm working on this for some days and stil dont fid a way. I'm trying to accoun the power consumption of the kernel reading the rapl counters in power/energy-cores/

19:41 <fche> some such counters can be most easily read from userspace

19:48 <silvaleandrodp> but is it possible to account all proccessing done in kernel space from user space?

19:50 <fche> that's a good question. have you been able to find a perf.* probe parametrization that at least attached to those rapl counters?

20:01 <silvaleandrodp> sorry i was disconnected, from terminal we can read it #perf -a -e power/energy-cores/

20:02 <silvaleandrodp> the configurations of this counter can be found at /sys/bus/event_source/devices/power

20:02 <silvaleandrodp> reading it i thik it would be perf.type(14).config(0)

20:04 <fche> hm on my machine here /sys/bus/event_source/devices/power/type:8 ... but the kernel rejects probe perf.type(8).config(0) {}

20:38 <silvaleandrodp> it's not working here too, in my system it is 14 and i'm almost sure tha it is the right value for type, but i dont know well what the valye for config

20:49 <fche> yeah, sorry I don't have a canned answer for you

20:49 <fche> will add a PR to track this at least

20:53 <fche> https://sourceware.org/bugzilla/show_bug.cgi?id=22289 <- cc self on this if able

21:47 <irker275> systemtap: dsmith systemtap.git:refs/heads/master * release-3.1-319-g3de3622 / buildrun.cxx tapset/linux/dev.stp tapset/linux/ioblock.stp testsuite/buildok/dev-embedded.stp: Fixed PR22287 by updating the ioblock tapset. http://tinyurl.com/y76mkzjj

21:49 <silvaleandrodp> thankyou so much for the support fche

21:49 <fche> no problem, I just wish I had a ready-made answer

21:50 <silvaleandrodp> if i find something useful i came to share

21:50 <fche> if we're lucky, some type/config tuple will do what we need; even if that number tuple is machine-variable (we can try dealing with that)

21:51 <fche> if we need -more- parameters for that perf_event_attr struct that defines the probe target, we can let the translator handle those

21:51 <fche> it may not be much work, but it's not something the team is going to do immediately

21:51 <fche> would be glad to help advise if someone else wants to try