Revert "KVM: Support vCPU-based gfn->hva cache"
authorPaolo Bonzini <pbonzini@redhat.com>
Tue, 2 May 2017 14:20:18 +0000 (16:20 +0200)
committerPaolo Bonzini <pbonzini@redhat.com>
Wed, 3 May 2017 14:30:26 +0000 (16:30 +0200)
commit4e335d9e7ddbcf83d03e7fbe65797ebed2272c18
treeef791b6623e77a8d3abef05ff0410709def329d8
parentee5f7d79a80550179b258417442b7bdbccaf476a
Revert "KVM: Support vCPU-based gfn->hva cache"

This reverts commit bbd6411513aa8ef3ea02abab61318daf87c1af1e.

I've been sitting on this revert for too long and it unfortunately
missed 4.11.  It's also the reason why I haven't merged ring-based
dirty tracking for 4.12.

Using kvm_vcpu_memslots in kvm_gfn_to_hva_cache_init and
kvm_vcpu_write_guest_offset_cached means that the MSR value can
now be used to access SMRAM, simply by making it point to an SMRAM
physical address.  This is problematic because it lets the guest
OS overwrite memory that it shouldn't be able to touch.

Cc: stable@vger.kernel.org
Fixes: bbd6411513aa8ef3ea02abab61318daf87c1af1e
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/lapic.c
arch/x86/kvm/x86.c
include/linux/kvm_host.h
virt/kvm/kvm_main.c