Dne 21.3.2012 06:31, Satheesh Rajendran napsal(a): > Lucas, > > cgroup memory_limit,memory_memsw_limit,memory_move test ran in the > rhel6.2ga causing a kernel panic in host and this is reproducible every > time. > > It is not capturing the test results due to the panic. > > I have added some pdb trace statements for debugging in the file > (/client/tests/kvm/tests/cgroup.py) attached the same here. > > I ve executed the test and captured the bash log(bash-log.txt) and var > log messages(log-message.txt) > > Though other cgroup tests are running fine which are using the same > function (cgroup.set_cgroup). > > Tried even to enable the core dump, not getting triggered. > > Please have look into this and comment. > > Let me know in case of further inputs needed. > > Regards, > -Satheesh.
Nice :-) Yes on my F16 it also works fine and I just checked on my RHEL6.2 it also crashes. There is one interesting part of your logs in file "kernel-panic-serial-log.png" where booth killed processes have the exact amount of total-vm memory. Looks like some uninitialized value / corruption. Anyway the test itself creates machine with 1GB memory and limits the maximum RSS usage to half. Than it tries to allocate 562MB using dd in guest. So when cgroups work it should end-up swapping and consuming only 512MB RSS memory. If cgroup limit won't work it should consume 562 + system memory (let's say 128MB). I tested it on 5GB ram machine so it was way beyond the machine memory limit. The problem is probably somewhere in cgroups / memory subsystem. Regards, Lukáš _______________________________________________ Autotest mailing list Autotest@test.kernel.org http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
