From: Brandeburg, J. <jes...@in...> - 2007-04-24 17:59:16
|
Maynard Johnson wrote: > Jesse, > Maybe someone on this list can offer an actual solution, but as for > troubleshooting the problem, I suggest running the opcontrol commands > one at a time, as follows: opcontrol --init > opcontrol --vmlinux=3Dblah > opcontrol --start-daemon > opcontrol --start Thank you for your suggestion, I tried this ordering, except I added opcontrol --start-daemon --verbose=3Dall I did opcontrol --start, saw that it claimed it started profiling in /var/lib/oprofile/samples/oprofiled.log I then did some kernel stuff (find /) and then executed opcontrol --dump This is what was printed in oprofiled.log: oprofiled started Tue Apr 24 10:19:15 2007 kernel pointer size: 8 Read buffer of 16314 entries. CPU_SWITCH to 1 Etc.... CPU_SWITCH to 0 Etc... That is all that is printed until I stop the daemon, when it prints: Tue Apr 24 10:29:15 2007 Nr. sample dumps: 2 ... all samples : 0 after this =20 > Maybe this will give you a better idea where/when things > start falling > apart. >=20 > BTW, have you tried running with SLES 10's kernel and the > oprofile rpm > that comes with the distribution? You'll find the oprofile > rpm on the SDK. Actually, yes, I have, but oprofile 0.9.1 included doesn't recognize my CPU and so won't work. Thanks for your help, any other ideas? I wonder if my kernel isn't configured correctly for some reason. It is x86_64, doesn't oprofile work there? Would it help if I attached my .config? |