On Mon, Nov 11, 2019 at 07:08:20PM +0100, Greg Kroah-Hartman wrote:
On Mon, Nov 11, 2019 at 09:57:19AM -0800, Sean Christopherson wrote:
On Mon, Nov 11, 2019 at 06:48:59PM +0100, Greg Kroah-Hartman wrote:
Thanks for figuring this out, can you send us a patch that we can apply to fix this issue in the stable tree?
Can do. A custom backport will be need for 4.20 and earlier, not 4.19 and earlier. I misremembered when we did the VMX refactoring.
For 5.0, 5.1 and 5.2, commit bf03d4f93347 can be applied directly.
5.0, 5.1, and 5.2 are all long end-of-life, they are not getting any updates and no one should be using them, so nothing to worry about there.
Backports sent for 4.14 and 4.19. 4.9 and 4.4 aren't affected as the bug was introduced in 4.14. by commit d1cd3ce90044 ("KVM: MMU: check guest CR3 reserved bits based on its physical address width.").
I did send patches for 4.9 and 4.4 for another PAE bug fix that I ran into while backporting; commit d35b34a9a70e ("kvm: mmu: Don't read PDPTEs when paging is not enabled"). I'm not aware of bug reports, but the patch is quite safe and should have been tagged for stable.