File 58343dc2-x86-hvm-Fix-the-handling-of-non-presen... of Package xen

Overview Repositories Revisions Requests Users Attributes Meta

File 58343dc2-x86-hvm-Fix-the-handling-of-non-present-segments.patch of Package xen (Revision fe1d1744a8ed7820af5694febca8c8fd)

Currently displaying revision fe1d1744a8ed7820af5694febca8c8fd , Show latest

Subject: x86/hvm: Fix the handling of non-present segments
From: Andrew Cooper andrew.cooper3@citrix.com Tue Nov 22 14:27:40 2016 +0100
Date: Tue Nov 22 14:27:40 2016 +0100:
Git: 877b7602876b54f2c7c4c19374ee189db6194d73

In 32bit, the data segments may be NULL to indicate that the segment is
ineligible for use.  In both 32bit and 64bit, the LDT selector may be NULL to
indicate that the entire LDT is ineligible for use.  However, nothing in Xen
actually checks for this condition when performing other segmentation
checks.  (Note however that limit and writeability checks are correctly
performed).

Neither Intel nor AMD specify the exact behaviour of loading a NULL segment.
Experimentally, AMD zeroes all attributes but leaves the base and limit
unmodified.  Intel zeroes the base, sets the limit to 0xfffffff and resets the
attributes to just .G and .D/B.

The use of the segment information in the VMCB/VMCS is equivalent to a native
pipeline interacting with the segment cache.  The present bit can therefore
have a subtly different meaning, and it is now cooked to uniformly indicate
whether the segment is usable or not.

GDTR and IDTR don't have access rights like the other segments, but for
consistency, they are treated as being present so no special casing is needed
elsewhere in the segmentation logic.

AMD hardware does not consider the present bit for %cs and %tr, and will
function as if they were present.  They are therefore unconditionally set to
present when reading information from the VMCB, to maintain the new meaning of
usability.

Intel hardware has a separate unusable bit in the VMCS segment attributes.
This bit is inverted and stored in the present field, so the hvm code can work
with architecturally-common state.

This is CVE-2016-9386 / XSA-191.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 04beafa8e6c66f5cd814c00e2d2b51cfbc41cb8a
master date: 2016-11-22 13:44:50 +0100

--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -2173,6 +2173,10 @@ int hvm_virtual_to_linear_addr(
          * COMPATIBILITY MODE: Apply segment checks and add base.
          */
 
+        /* Segment not valid for use (cooked meaning of .p)? */
+        if ( !reg->attr.fields.p )
+            return 0;
+
         switch ( access_type )
         {
         case hvm_access_read:
@@ -2370,6 +2374,10 @@ static int hvm_load_segment_selector(
     hvm_get_segment_register(
         v, (sel & 4) ? x86_seg_ldtr : x86_seg_gdtr, &desctab);
 
+    /* Segment not valid for use (cooked meaning of .p)? */
+    if ( !desctab.attr.fields.p )
+        goto fail;
+
     /* Check against descriptor table limit. */
     if ( ((sel & 0xfff8) + 7) > desctab.limit )
         goto fail;
--- a/xen/arch/x86/hvm/svm/svm.c
+++ b/xen/arch/x86/hvm/svm/svm.c
@@ -523,6 +523,7 @@ static void svm_get_segment_register(str
     {
     case x86_seg_cs:
         memcpy(reg, &vmcb->cs, sizeof(*reg));
+        reg->attr.fields.p = 1;
         reg->attr.fields.g = reg->limit > 0xFFFFF;
         break;
     case x86_seg_ds:
@@ -556,13 +557,16 @@ static void svm_get_segment_register(str
     case x86_seg_tr:
         svm_sync_vmcb(v);
         memcpy(reg, &vmcb->tr, sizeof(*reg));
+        reg->attr.fields.p = 1;
         reg->attr.fields.type |= 0x2;
         break;
     case x86_seg_gdtr:
         memcpy(reg, &vmcb->gdtr, sizeof(*reg));
+        reg->attr.bytes = 0x80;
         break;
     case x86_seg_idtr:
         memcpy(reg, &vmcb->idtr, sizeof(*reg));
+        reg->attr.bytes = 0x80;
         break;
     case x86_seg_ldtr:
         svm_sync_vmcb(v);
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -774,10 +774,12 @@ void vmx_get_segment_register(struct vcp
     reg->sel = sel;
     reg->limit = limit;
 
-    reg->attr.bytes = (attr & 0xff) | ((attr >> 4) & 0xf00);
-    /* Unusable flag is folded into Present flag. */
-    if ( attr & (1u<<16) )
-        reg->attr.fields.p = 0;
+    /*
+     * Fold VT-x representation into Xen's representation.  The Present bit is
+     * unconditionally set to the inverse of unusable.
+     */
+    reg->attr.bytes =
+        (!(attr & (1u << 16)) << 7) | (attr & 0x7f) | ((attr >> 4) & 0xf00);
 
     /* Adjust for virtual 8086 mode */
     if ( v->arch.hvm_vmx.vmx_realmode && seg <= x86_seg_tr 
@@ -857,11 +859,11 @@ static void vmx_set_segment_register(str
         }
     }
 
-    attr = ((attr & 0xf00) << 4) | (attr & 0xff);
-
-    /* Not-present must mean unusable. */
-    if ( !reg->attr.fields.p )
-        attr |= (1u << 16);
+    /*
+     * Unfold Xen representation into VT-x representation.  The unusable bit
+     * is unconditionally set to the inverse of present.
+     */
+    attr = (!(attr & (1u << 7)) << 16) | ((attr & 0xf00) << 4) | (attr & 0xff);
 
     /* VMX has strict consistency requirement for flag G. */
     attr |= !!(limit >> 20) << 15;
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -1210,6 +1210,10 @@ protmode_load_seg(
                                  &desctab, ctxt)) )
         return rc;
 
+    /* Segment not valid for use (cooked meaning of .p)? */
+    if ( !desctab.attr.fields.p )
+        goto raise_exn;
+
     /* Check against descriptor table limit. */
     if ( ((sel & 0xfff8) + 7) > desctab.limit )
         goto raise_exn;

Places

File 58343dc2-x86-hvm-Fix-the-handling-of-non-present-segments.patch of Package xen (Revision fe1d1744a8ed7820af5694febca8c8fd)

Places