CentOS

It's interesting that kernel-ml did not work for you. The referenced patch is commit 7b0501b1e7cddd32b265178e32d332bdfbb532d4 of the upstream (kernel.org) tree. It first appeared in kernel 4.8-rc2. This means that the patch is in kernel-ml which is now at 4.11.

Did I understood correctly that fresh install was done and then cpu added?

Have you tried doing after the addition a fresh install and see if the issue repeats. For example with certain servers changing from UEFi to Bios boot would require fresh install and not sure if all BIOS/HW changes are recommended without new install.

Redhat confirmed bug still in 7.3 that has already been fixed for 7.4 and has been backported to the z-stream/EUS kernel kernel-3.10.0-327.55.1 .

I tested kernel-3.10.0-327.55.1 and sure enough it works.

I'm still following up with Redhat to confirm that UEFI is fixed as well.

toracat wrote:It's interesting that kernel-ml did not work for you. The referenced patch is commit 7b0501b1e7cddd32b265178e32d332bdfbb532d4 of the upstream (kernel.org) tree. It first appeared in kernel 4.8-rc2. This means that the patch is in kernel-ml which is now at 4.11.

Yes that is interesting. Just to make sure I wasn't kidding myself I verified kernel-ml-4.11.6-1.el7.elrepo.x86_64 is broken once more. Looking at the kernel changelog there have been numerous changes related to this so perhaps there was some refactoring required since the original patch that Redhat intends to eventually upload to the mainstream kernel.

# rpm -q --changelog kernel | grep -i max_logical_packages
- [x86] smp: Fix __max_logical_packages value setup (Prarit Bhargava) [1398173 1394239]
- [x86] revert "smp: Fix __max_logical_packages value setup" (Prarit Bhargava) [1398173 1394239]
- [x86] smp: Fix __max_logical_packages value setup (Frank Ramsay) [1358312]

Matt