KVM: avoid double put_page with gfn-to-pfn cache

If the cache's user host virtual address becomes invalid, there
is still a path from kvm_gfn_to_pfn_cache_refresh() where __release_gpc()
could release the pfn but the gpc->pfn field has not been overwritten
with an error value.  If this happens, kvm_gfn_to_pfn_cache_unmap will
call put_page again on the same page.

Cc: stable@vger.kernel.org
Fixes: 982ed0de4753 ("KVM: Reinstate gfn_to_pfn_cache with invalidation support")
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

diff --git a/virt/kvm/pfncache.c b/virt/kvm/pfncache.c
index ce878f4be4daab2dc5e1677b2a3c3edd760959e9..1621f8efd9616a1c3aa406a54fe98f27aeb47831 100644 (file)
--- a/virt/kvm/pfncache.c
+++ b/virt/kvm/pfncache.c
@@ -191,6 +191,7 @@ int kvm_gfn_to_pfn_cache_refresh(struct kvm *kvm, struct gfn_to_pfn_cache *gpc,
                gpc->uhva = gfn_to_hva_memslot(gpc->memslot, gfn);
 
                if (kvm_is_error_hva(gpc->uhva)) {
+                       gpc->pfn = KVM_PFN_ERR_FAULT;
                        ret = -EFAULT;
                        goto out;
                }