KVM: x86: Add memory barrier on vmcs field lookup
authorAndrew Honig <ahonig@google.com>
Wed, 10 Jan 2018 18:12:03 +0000 (10:12 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 17 Jan 2018 08:45:24 +0000 (09:45 +0100)
commit 75f139aaf896d6fdeec2e468ddfa4b2fe469bf40 upstream.

This adds a memory barrier when performing a lookup into
the vmcs_field_to_offset_table.  This is related to
CVE-2017-5753.

Signed-off-by: Andrew Honig <ahonig@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
arch/x86/kvm/vmx.c

index 93de008..e6f291d 100644 (file)
@@ -888,8 +888,16 @@ static inline short vmcs_field_to_offset(unsigned long field)
 {
        BUILD_BUG_ON(ARRAY_SIZE(vmcs_field_to_offset_table) > SHRT_MAX);
 
-       if (field >= ARRAY_SIZE(vmcs_field_to_offset_table) ||
-           vmcs_field_to_offset_table[field] == 0)
+       if (field >= ARRAY_SIZE(vmcs_field_to_offset_table))
+               return -ENOENT;
+
+       /*
+        * FIXME: Mitigation for CVE-2017-5753.  To be replaced with a
+        * generic mechanism.
+        */
+       asm("lfence");
+
+       if (vmcs_field_to_offset_table[field] == 0)
                return -ENOENT;
 
        return vmcs_field_to_offset_table[field];