KVM: SVM: do not zero out segment attributes if segment is unusable or not present
authorRoman Pen <roman.penyaev@profitbricks.com>
Thu, 1 Jun 2017 08:55:03 +0000 (10:55 +0200)
committerPaolo Bonzini <pbonzini@redhat.com>
Thu, 1 Jun 2017 09:21:17 +0000 (11:21 +0200)
This is a fix for the problem [1], where VMCB.CPL was set to 0 and interrupt
was taken on userspace stack.  The root cause lies in the specific AMD CPU
behaviour which manifests itself as unusable segment attributes on SYSRET.
The corresponding work around for the kernel is the following:

61f01dd941ba ("x86_64, asm: Work around AMD SYSRET SS descriptor attribute issue")

In other turn virtualization side treated unusable segment incorrectly and
restored CPL from SS attributes, which were zeroed out few lines above.

In current patch it is assured only that P bit is cleared in VMCB.save state
and segment attributes are not zeroed out if segment is not presented or is
unusable, therefore CPL can be safely restored from DPL field.

This is only one part of the fix, since QEMU side should be fixed accordingly
not to zero out attributes on its side.  Corresponding patch will follow.

[1] Message id: CAJrWOzD6Xq==b-zYCDdFLgSRMPM-NkNuTSDFEtX=7MreT45i7Q@mail.gmail.com

Signed-off-by: Roman Pen <roman.penyaev@profitbricks.com>
Signed-off-by: Mikhail Sennikovskii <mikhail.sennikovskii@profitbricks.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim KrÄmář <rkrcmar@redhat.com>
Cc: kvm@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/svm.c

index a654372efea183e49ab5992f3fff7c051b7a72db..ba9891ac5c568f1798555bfa9dcbc421fff5ae2a 100644 (file)
@@ -1840,6 +1840,7 @@ static void svm_get_segment(struct kvm_vcpu *vcpu,
                 */
                if (var->unusable)
                        var->db = 0;
+               /* This is symmetric with svm_set_segment() */
                var->dpl = to_svm(vcpu)->vmcb->save.cpl;
                break;
        }
@@ -1980,18 +1981,14 @@ static void svm_set_segment(struct kvm_vcpu *vcpu,
        s->base = var->base;
        s->limit = var->limit;
        s->selector = var->selector;
-       if (var->unusable)
-               s->attrib = 0;
-       else {
-               s->attrib = (var->type & SVM_SELECTOR_TYPE_MASK);
-               s->attrib |= (var->s & 1) << SVM_SELECTOR_S_SHIFT;
-               s->attrib |= (var->dpl & 3) << SVM_SELECTOR_DPL_SHIFT;
-               s->attrib |= (var->present & 1) << SVM_SELECTOR_P_SHIFT;
-               s->attrib |= (var->avl & 1) << SVM_SELECTOR_AVL_SHIFT;
-               s->attrib |= (var->l & 1) << SVM_SELECTOR_L_SHIFT;
-               s->attrib |= (var->db & 1) << SVM_SELECTOR_DB_SHIFT;
-               s->attrib |= (var->g & 1) << SVM_SELECTOR_G_SHIFT;
-       }
+       s->attrib = (var->type & SVM_SELECTOR_TYPE_MASK);
+       s->attrib |= (var->s & 1) << SVM_SELECTOR_S_SHIFT;
+       s->attrib |= (var->dpl & 3) << SVM_SELECTOR_DPL_SHIFT;
+       s->attrib |= ((var->present & 1) && !var->unusable) << SVM_SELECTOR_P_SHIFT;
+       s->attrib |= (var->avl & 1) << SVM_SELECTOR_AVL_SHIFT;
+       s->attrib |= (var->l & 1) << SVM_SELECTOR_L_SHIFT;
+       s->attrib |= (var->db & 1) << SVM_SELECTOR_DB_SHIFT;
+       s->attrib |= (var->g & 1) << SVM_SELECTOR_G_SHIFT;
 
        /*
         * This is always accurate, except if SYSRET returned to a segment
@@ -2000,7 +1997,8 @@ static void svm_set_segment(struct kvm_vcpu *vcpu,
         * would entail passing the CPL to userspace and back.
         */
        if (seg == VCPU_SREG_SS)
-               svm->vmcb->save.cpl = (s->attrib >> SVM_SELECTOR_DPL_SHIFT) & 3;
+               /* This is symmetric with svm_get_segment() */
+               svm->vmcb->save.cpl = (var->dpl & 3);
 
        mark_dirty(svm->vmcb, VMCB_SEG);
 }