07:22 <vbernat> hey! is "probe kernel.function("context_switch@/build/linux-wxhyRy/linux-4.3.3/kernel/sched/core.c:2618") (address 0xffffffff81581992) registration error (rc -22)" a known bug when using scheduler.cpu_off/cpu_on?

07:23 <vbernat> I did notice #16764 but since this happens when using provided probe function, maybe it's a new thing

07:23 <vbernat> systemtap 2.9

07:24 <vbernat> and Debian kernel

07:57 <jvv> this seems like a nice source of info http://myaut.github.io/dtrace-stap-book/

12:35 <sjas> jvv: thank you very much

13:28 <fche> vbernat, it's not a bug exactly; it's what the kernel refuses to do for us

13:29 <fche> wonder why the kernel.trace("sched_switch") preferred alternative is not resolving; is your kernel built without tracepoint support?

13:29 <vbernat> fche: CONFIG_TRACEPOINTS=y

13:30 <fche> stap -L 'kernel.trace("sched_switch")' ?

13:31 <vbernat> kernel.trace("sched:sched_switch") $prev:struct task_struct* $next:struct task_struct*

13:31 <vbernat>

13:31 <fche> hm, so probe scheduler.cpu_off { } should prefer that, and not do the kernel.function() business

13:32 <vbernat> humm, the problem seem to be on my side

13:32 <vbernat> on my test machine with the same kernel, I don't get anything with stap -L

13:34 <fche> maybe missing the package that supplies kernel header files ?

13:34 <vbernat> fche: I shouldn't need it for tracepoints?

13:34 <vbernat> but in fact, I am using the exact FS between the two machines

13:34 <fche> stap computes tracepoints based on headers

13:35 <vbernat> one is a QEmu virtual machine using the original FS mounted over 9P

13:35 <fche> maybe they're not running the same kernel version?

13:35 <vbernat> yes, they are not, a small update difference

13:35 <fche> anyway that's where I'd start looking (stap --vp 08 -e 'probe kernel.trace("*") {}' ... see how trace* files are processed)

13:36 <vbernat> thanks, I'll investigate. My setup being a bit special, it's likely that I have messed something.

13:37 <fche> would be interesting to find out why stap doesn't print diagnostics for your case (assuming it doesn't)

14:32 <irker414> systemtap: jlebon systemtap.git:refs/heads/master * release-2.9-221-g5fa1d0a / tapsets.cxx: tapsets.cxx: fix indentation to appease GCC 6 http://tinyurl.com/j2qossm

14:33 <fche> "peace in our time"

14:33 <jlebon> mjw: ^ should be gcc 6 ready now :)

14:33 <jlebon> fche: haha

14:33 <mjw> nice

14:33 <mjw> will try it "soon"

14:34 <mjw> but I promised valgrind first :)

14:35 <jlebon> hehe

14:35 <jlebon> error printing is super nice overall now

14:35 <fche> in gcc? yeah, some awesome work there recently

14:36 <fche> competition from llvm motivating people

14:36 <jlebon> heh, yeah in gcc. the underlining and coloring and everything.

14:36 <jlebon> yeah i suppose so. yay competition \o/

16:12 <irker414> systemtap: mcermak systemtap.git:refs/heads/master * release-2.9-222-gd679d19 / configure.ac man/Makefile.am man/cs/Makefile.am systemtap.spec: Preparation for the czech localization of manpages. http://tinyurl.com/hjchgpb

16:12 <irker414> systemtap: mcermak systemtap.git:refs/heads/master * release-2.9-223-gf0cf22c / man/cs/stap.1 man/cs/stapprobes.3stap: Add czech version of stap.1 and stapprobes.3stap manpages. http://tinyurl.com/zzvn5g6

16:12 <irker414> systemtap: mcermak systemtap.git:refs/heads/master * release-2.9-224-g7ba226f / Makefile.in aclocal.m4 configure doc/Makefile.in doc/SystemTap_Tapset_Reference/Makefile.in doc/beginners/Makefile.in java/Makefile.in man/Makefile.in man/cs/Makefile.in stapdyn/Makefile.in staprun/Makefile.in: Autoreconf after adding czech manpages. http://tinyurl.com/hk7r6ua

16:12 <fche> thanks martin!

19:28 <vbernat> fche: in fact, removing .systemtap solved the problem

19:28 <vbernat> it is shared between my VM and the host

19:28 <vbernat> so I suppose that because of the kernel mismatch, there were a problem

19:30 <fche> interesting; stap tries to protect itself against that by including kernel version numbers etc. in the hashes used in the cache

19:30 <fche> how does 'uname -r' and 'uname -a' vary between the two?

19:55 <irker414> systemtap: dsmith systemtap.git:refs/heads/master * release-2.9-225-g9fef07f / elaborate.cxx parse.cxx staptree.cxx staptree.h testsuite/systemtap.pass1-4/parse-semok.exp: PR19510 partial fix - improve '-p1' output. http://tinyurl.com/jq3htk3

22:14 <irker414> systemtap: dsmith systemtap.git:refs/heads/master * release-2.9-226-gd014a4e / testsuite/systemtap.base/utrace_syscall_args.stp: Fix systemtap.base/utrace_syscall_args.stp for s390x kernels >= 2.6.33. http://tinyurl.com/jhmzjmx