On Thu, Sep 12, 2019 at 02:20:09PM -0700, Jim Mattson wrote:
> On Wed, Sep 11, 2019 at 9:18 PM Fuqian Huang <huangfq.dax...@gmail.com> wrote:
> >
> > Emulation of VMPTRST can incorrectly inject a page fault
> > when passed an operand that points to an MMIO address.
> > The page fault will use uninitialized kernel stack memory
> > as the CR2 and error code.
> >
> > The right behavior would be to abort the VM with a KVM_EXIT_INTERNAL_ERROR
> > exit to userspace; however, it is not an easy fix, so for now just ensure
> > that the error code and CR2 are zero.
> >
> > Signed-off-by: Fuqian Huang <huangfq.dax...@gmail.com>
> > ---
> > arch/x86/kvm/x86.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index 290c3c3efb87..7f442d710858 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -5312,6 +5312,7 @@ int kvm_write_guest_virt_system(struct kvm_vcpu
> > *vcpu, gva_t addr, void *val,
> > /* kvm_write_guest_virt_system can pull in tons of pages. */
> > vcpu->arch.l1tf_flush_l1d = true;
> >
> > + memset(exception, 0, sizeof(*exception));
> > return kvm_write_guest_virt_helper(addr, val, bytes, vcpu,
> > PFERR_WRITE_MASK, exception);
> > }
> > --
> > 2.11.0
> >
> Perhaps you could also add a comment like the one Paolo added when he
> made the same change in kvm_read_guest_virt?
> See commit 353c0956a618 ("KVM: x86: work around leak of uninitialized
> stack contents (CVE-2019-7222)").
I have a better hack-a-fix, we can handle the unexpected MMIO using master
abort semantics, i.e. reads return all ones, writes are dropped. It's not
100% correct as KVM won't handle the case where the address is legit MMIO,
but it's at least sometimes correct and thus better than a #PF.
Patch and a unit test incoming...
