On Tue, 16 Jan 2018, Ingo Molnar wrote:
Thanks for the confirmation. Now the fix is upstream and Greg queued it up as well earlier today, so this fix is part of the next -stable iteration as well.
Greg, please note that you need this only for the 4.14-stable.
4.4-stable and 4.9-stable already do the right thing (always or-ing 0 into CR3 on non-PCID systems) due to the way how x86_cr3_pcid_user handling is done.