KVM: x86: apply kvmclock offset to guest wall clock time
authorBruce Rogers <brogers@suse.com>
Fri, 20 Jul 2012 16:44:24 +0000 (10:44 -0600)
committerMarcelo Tosatti <mtosatti@redhat.com>
Wed, 1 Aug 2012 20:23:50 +0000 (17:23 -0300)
When a guest migrates to a new host, the system time difference from the
previous host is used in the updates to the kvmclock system time visible
to the guest, resulting in a continuation of correct kvmclock based guest
timekeeping.

The wall clock component of the kvmclock provided time is currently not
updated with this same time offset. Since the Linux guest caches the
wall clock based time, this discrepency is not noticed until the guest is
rebooted. After reboot the guest's time calculations are off.

This patch adjusts the wall clock by the kvmclock_offset, resulting in
correct guest time after a reboot.

Cc: Zachary Amsden <zamsden@gmail.com>
Signed-off-by: Bruce Rogers <brogers@suse.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
arch/x86/kvm/x86.c

index 59b59508ff07dc2623887d6be43859215d5bd589..42bce48f692850cf3cadf96e83c86e5f8bf760ee 100644 (file)
@@ -925,6 +925,10 @@ static void kvm_write_wall_clock(struct kvm *kvm, gpa_t wall_clock)
         */
        getboottime(&boot);
 
+       if (kvm->arch.kvmclock_offset) {
+               struct timespec ts = ns_to_timespec(kvm->arch.kvmclock_offset);
+               boot = timespec_sub(boot, ts);
+       }
        wc.sec = boot.tv_sec;
        wc.nsec = boot.tv_nsec;
        wc.version = version;