The hypervisor code for wallclock setting in pvclock is subject to race, since the version variable is static, and thus, shared among all guests. This means that theoretically, one guest can mess up with the others view of version, causing it to be different among reads, which can lead to a livelock.
Verify this issue on : # uname -r 2.6.18-229.el5 # rpm -q kvm kvm-83-207.el5 steps:,start 7 guest on 8cores host with -no-kvm-pit-reinjection -rtc-td-hack -startdate now actual results:all guests didn't observe large offset(all less than 3s) after boot,pass. this issue has already been fixed.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2011-0028.html