KVM: MMU: prepopulate guest pages after write-protecting

Zdenek reported a bug where a looping "dmsetup status" eventually hangs
on SMP guests.

The problem is that kvm_mmu_get_page() prepopulates the shadow MMU
before write protecting the guest page tables. By doing so, it leaves a
window open where the guest can mark a pte as present while the host has
shadow cached such pte as "notrap". Accesses to such address will fault
in the guest without the host having a chance to fix the situation.

Fix by moving the write protection before the pte prefetch.
Signed-off-by: NMarcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: NAvi Kivity <avi@qumranet.com>

arch/x86/kvm/mmu.c

@@ -852,9 +852,9 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
 	sp->gfn = gfn;
 	sp->role = role;
 	hlist_add_head(&sp->hash_link, bucket);
-	vcpu->arch.mmu.prefetch_page(vcpu, sp);
 	if (!metaphysical)
 		rmap_write_protect(vcpu->kvm, gfn);
+	vcpu->arch.mmu.prefetch_page(vcpu, sp);
 	return sp;
 }