Comment # 8 on bug 1087490 from Ralf K�lmel

Created attachment 766234 [details]
msr-mod-bits output

After following your steps (activating MWAIT, C6, running your script) the
tests are running into CPU stucks with the result that the system is totally
blocked.
Many console messages are given in the manner of :

Message from syslogd@compute4 at Apr  6 02:37:12 ...
 kernel:[ 5466.403421] watchdog: BUG: soft lockup - CPU#67 stuck for 22s!
[stress-ng-numa:36806            
Message from syslogd@compute4 at Apr  6 02:37:12 ...
 kernel:[ 5466.947425] watchdog: BUG: soft lockup - CPU#84 stuck for 22s!
[kworker/84:1:45834]               
Message from syslogd@compute4 at Apr  6 02:37:40 ...
 kernel:[ 5494.403509] watchdog: BUG: soft lockup - CPU#67 stuck for 23s!
[stress-ng-numa:36806]
Message from syslogd@compute4 at Apr  6 02:37:40 ...
 kernel:[ 5494.947510] watchdog: BUG: soft lockup - CPU#84 stuck for 22s!
[kworker/84:1:45834]

I've searched after MCE in the logs of the past 2 months, but i have seen only
MCE  initialization and during test on one day MCEs about corrupted pages in
the manner of :

/var/log/messages-20180202.xz:2018-02-01T01:03:43.074004+01:00 compute4 kernel:
[46702.580100] MCE 0x23885a6: corrupted page was clean: dropped without side
effects
/var/log/messages-20180202.xz:2018-02-01T01:03:43.074005+01:00 compute4 kernel:
[46702.580114] MCE 0x23885a6: recovery action for clean LRU page: Recovered