Description of problem: In order get migration of kvm guests working on remote and unsynchronized hosts, we need two new ioctls that will allow us to adjust pvclock value. Those patches are in upstream kvm kernel, and I will provide a backport shortly.
scratch build at : http://koji.fedoraproject.org/koji/taskinfo?taskID=1763328
Created attachment 365834 [details] Patch for it.
Okay, it's queued up in Avi's tree here: http://git.kernel.org/?p=linux/kernel/git/avi/kvm.git;a=commitdiff;h=a6048dd2da http://git.kernel.org/?p=linux/kernel/git/avi/kvm.git;a=commitdiff;h=958b0c5497 which means it should be in 2.6.33 The qemu-kvm side is here: http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=commitdiff;h=e4d6d49061 http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=commitdiff;h=11ed4b344c http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=commitdiff;h=79529cd015 which should be in qemu-kvm-0.12.0 Glauber: are you proposing to pull the new ioctl for 2.6.31 in Fedora 12? What are the symptoms of the bug? i.e. how bad is it? could it wait until Fedora 13?
(In reply to comment #3) > Okay, it's queued up in Avi's tree here: > > > http://git.kernel.org/?p=linux/kernel/git/avi/kvm.git;a=commitdiff;h=a6048dd2da > > http://git.kernel.org/?p=linux/kernel/git/avi/kvm.git;a=commitdiff;h=958b0c5497 > > which means it should be in 2.6.33 > > The qemu-kvm side is here: > > http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=commitdiff;h=e4d6d49061 > http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=commitdiff;h=11ed4b344c > http://git.kernel.org/?p=virt/kvm/qemu-kvm.git;a=commitdiff;h=79529cd015 > > which should be in qemu-kvm-0.12.0 > > Glauber: are you proposing to pull the new ioctl for 2.6.31 in Fedora 12? I guess we could. The patch is there, and is already a backport to F12 tree, should apply just fine. So it's really up to the kernel maintainer. I can backport the userspace part if we get the kernel part working. > What are the symptoms of the bug? i.e. how bad is it? could it wait until > Fedora 13? Imagine two hosts A and B A has a smaller monotonic clock than B (say, B was booted first) If you migrate A -> B, you get a time jump. If you migrate B -> A, you get backwards timing, and machine hangs
(In reply to comment #4) > (In reply to comment #3) > > Okay, it's queued up in Avi's tree here: > > > > > > http://git.kernel.org/?p=linux/kernel/git/avi/kvm.git;a=commitdiff;h=a6048dd2da > > > > http://git.kernel.org/?p=linux/kernel/git/avi/kvm.git;a=commitdiff;h=958b0c5497 > > > > which means it should be in 2.6.33 ... > > Glauber: are you proposing to pull the new ioctl for 2.6.31 in Fedora 12? > > I guess we could. The patch is there, and is already a backport to F12 tree, > should apply just fine. Ah, good - the patch you attached is the combination of both those commits. Justin, it sounds important enough to get this queued up for a post-GA update.
This bug appears to have been reported against 'rawhide' during the Fedora 13 development cycle. Changing version to '13'. More information and reason for this action is here: http://fedoraproject.org/wiki/BugZappers/HouseKeeping
F13 now has kernel 2.6.34