From d1f32e86ad69e6b39635807c71fb954a8a31172b Mon Sep 17 00:00:00 2001 From: Glauber Costa <glommer@redhat.com> Date: Fri, 3 Dec 2010 14:57:35 -0200 Subject: [PATCH 2/2] load registers after restoring pvclock msrs RH-Author: Glauber Costa <glommer@redhat.com> Message-id: <patch-14173-clone-for-rhel55-rhel55> Patchwork-id: 14287 O-Subject: [RHEL5.6 PATCH] load registers after restoring pvclock msrs Bugzilla: 660239 RH-Acked-by: Eduardo Habkost <ehabkost@redhat.com> RH-Acked-by: Juan Quintela <quintela@redhat.com> RH-Acked-by: Jes Sorensen <Jes.Sorensen@redhat.com> BZ: 655990 Upstream-status: N/A After we grab the pvclock-related values from the migration stream, we have to call kvm_load_registers() to call the kernel and make the changes visible to it. The absense of it makes migration with pvclock broken even with -M rhel5.5.0, which was supposed to work after BZ531701. Since BZ531701 did fix the problem, and was verified by QE and us, I can only speculate that we used to have a call to kvm_load_registers() being called as a side-effect of somethine else, that was removed in a recent patch. Signed-off-by: Glauber Costa <glommer@redhat.com> --- qemu/target-i386/machine.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) Signed-off-by: Eduardo Habkost <ehabkost@redhat.com> --- qemu/target-i386/machine.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/qemu/target-i386/machine.c b/qemu/target-i386/machine.c index a65b55d..f4d45a7 100644 --- a/qemu/target-i386/machine.c +++ b/qemu/target-i386/machine.c @@ -353,6 +353,7 @@ int cpu_load(QEMUFile *f, void *opaque, int version_id) if (version_id >= 8) { qemu_get_be64s(f, &env->system_time_msr); qemu_get_be64s(f, &env->wall_clock_msr); + kvm_load_registers(env); } if (version_id >= 9) { qemu_get_8s(f, &env->soft_interrupt); -- 1.7.3.2