CentOS

I have a couple of Virtual private servers, one running under Xen and the other KVM.

After installing the latest updates (including the latest kernel), neither boots. The virtual serial consoles provided by the VM services show nothing.

I also have a number of Centos 6.9 virtual machines (running under Xenserver 7.1) and after updating these fail to boot as well. I have reinstalled grub and they still fail. They reach the point where they would boot and then almost immediately shut down.

Listening in on chat in the CentOS-virt channel on freenode implies that the problem could be down to an incompatibility with VMs running in PV (paravirtualised) mode. Apparently, converting them to HVM (hardware assisted virtualisation) may work. I am checking this out at the moment and will post more if I find anything.

For the time being, I have re-enabled the original kernel, which is probably less than ideal given the latest vulnerabilities.

I can confirm the issue. All Centos 6.9 with latest kernel running under xenserver 7.x are affected. DO NOT UPDATE !
Let's wait for a fix, hopefully soon

This issue is apparently something to do with the new code for the "meltdown" vulnerability and is currently failing on both el6 and el7 latest kernels.

This is being tracked in https://bugs.centos.org/view.php?id=14336

Don't these kernel updates get any testing? I realize that CentOS may have very limited resources for testing, but doesn't Red Hat test these updates?

This isn't some subtle bug. This is a catastrophic bug.

Bug is already under investigation by Karanbir Singh. Centos Project Leader.

Let's keep finger crossed

I don't think RH care much about xen. The initial 7.4 kernel fails to boot as a xen guest and they've not scheduled a fix until 7.5, ETA unknown.

Does the new 6.9 kernel work under xen if you append pti=off (or maybe nopti) to the kernel command line?

The problem with kvm is most likely different as I have successfully installed the new 6.9 kernel on both kvm hosts and guests without problem (other than the expected eating of cpu).

I haven't the possibility of trying the "nopti" option at boot time right now, sorry.

Having the same issue on AWS using standard CentOS 6.9 PV minimal image. Looks like adding pti=off OR nopti to the 18.7 (Intel patch) kernel command line does not result in successful boot after yum update.