My kernel does have perf_event support in it. Debugging a liitle more, I found that the problem is in pfm_perf_event_os_detect function on pfmlib_perf_event.c file.
Since it defines PERF_PROC_FILE as "/proc/sys/kernel/perf_event_paranoid", the function is not able to access the file on kernel 2.6.31, since it does not exists. My workaround was to change #define PERF_PROC_FILE "/proc/sys/kernel/perf_event_paranoid" to #define PERF_PROC_FILE "/proc/sys/kernel/perf_counter_paranoid" and everything is working now. I'm just curious to know why you handle the "old" kernel version (2.6.31) in pfm_perf_detect function on pfmlib_perf_event_pmu.c file, but doesn't on pfmlib_perf_event.c. On Tue, Jul 5, 2011 at 10:25 PM, Vince Weaver <vweav...@eecs.utk.edu> wrote: > On Tue, 5 Jul 2011, Leonardo Piga wrote: > >> I'm using Ubuntu 9.10 x86 on a Intel Core2 Duo CPU Model: T9400, >> kernel 2.6.31-19. >> I also tried on other computer "Intel Core i7 CPU Model: 860" with >> Ubuntu 10.04 and I got the same message. > > are you sure your kernel has perf_event support in it? > > You'd be much better off with a newer kernel. perf_event support was > only introduced in 2.6.31 and it was relatively buggy at first, and > probably doesn't support your i7 machine. > > you can look for the file > /proc/sys/kernel/perf_event_paranoid > on a 2.6.32+ kernel or > /proc/sys/kernel/perf_counter_paranoid > on 2.6.31 to see if perf_event support was compiled in. > > Vince > -- Leonardo ------------------------------------------------------------------------------ All of the data generated in your IT infrastructure is seriously valuable. Why? It contains a definitive record of application performance, security threats, fraudulent activity, and more. Splunk takes this data and makes sense of it. IT sense. And common sense. http://p.sf.net/sfu/splunk-d2d-c2 _______________________________________________ perfmon2-devel mailing list perfmon2-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
