x86/shadow: Don't create self-linear shadow mappings for 4-level translated guests

When initially creating a monitor table for 4-level translated guests, don't
install a shadow-linear mapping.  This mapping is actually self-linear, and
trips up the writeable heuristic logic into following Xen's mappings, not the
guests' shadows it was expecting to follow.

A consequence of this is that sh_guess_wrmap() needs to cope with there being
no shadow-linear mapping present, which in practice occurs once each time a
vcpu switches to 4-level paging from a different paging mode.

An appropriate shadow-linear slot will be inserted into the monitor table
either while constructing lower level monitor tables, or by sh_update_cr3().

While fixing this, clarify the safety of the other mappings.  Despite
appearing unsafe, it is correct to create a guest-linear mapping for
translated domains; this is self-linear and doesn't point into the translated
domain.  Drop a dead clause for translate != external guests.

This is XSA-243.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Tim Deegan <tim@xen.org>

xen/arch/x86/mm/shadow/multi.c

@@ -1485,26 +1485,38 @@ void sh_install_xen_entries_in_l4(struct domain *d, mfn_t gl4mfn, mfn_t sl4mfn)
         sl4e[shadow_l4_table_offset(RO_MPT_VIRT_START)] = shadow_l4e_empty();
     }
 
-    /* Shadow linear mapping for 4-level shadows.  N.B. for 3-level
-     * shadows on 64-bit xen, this linear mapping is later replaced by the
-     * monitor pagetable structure, which is built in make_monitor_table
-     * and maintained by sh_update_linear_entries. */
-    sl4e[shadow_l4_table_offset(SH_LINEAR_PT_VIRT_START)] =
-        shadow_l4e_from_mfn(sl4mfn, __PAGE_HYPERVISOR_RW);
-
-    /* Self linear mapping.  */
-    if ( shadow_mode_translate(d) && !shadow_mode_external(d) )
+    /*
+     * Linear mapping slots:
+     *
+     * Calling this function with gl4mfn == sl4mfn is used to construct a
+     * monitor table for translated domains.  In this case, gl4mfn forms the
+     * self-linear mapping (i.e. not pointing into the translated domain), and
+     * the shadow-linear slot is skipped.  The shadow-linear slot is either
+     * filled when constructing lower level monitor tables, or via
+     * sh_update_cr3() for 4-level guests.
+     *
+     * Calling this function with gl4mfn != sl4mfn is used for non-translated
+     * guests, where the shadow-linear slot is actually self-linear, and the
+     * guest-linear slot points into the guests view of its pagetables.
+     */
+    if ( shadow_mode_translate(d) )
     {
-        // linear tables may not be used with translated PV guests
-        sl4e[shadow_l4_table_offset(LINEAR_PT_VIRT_START)] =
+        ASSERT(mfn_eq(gl4mfn, sl4mfn));
+
+        sl4e[shadow_l4_table_offset(SH_LINEAR_PT_VIRT_START)] =
             shadow_l4e_empty();
     }
     else
     {
-        sl4e[shadow_l4_table_offset(LINEAR_PT_VIRT_START)] =
-            shadow_l4e_from_mfn(gl4mfn, __PAGE_HYPERVISOR_RW);
+        ASSERT(!mfn_eq(gl4mfn, sl4mfn));
+
+        sl4e[shadow_l4_table_offset(SH_LINEAR_PT_VIRT_START)] =
+            shadow_l4e_from_mfn(sl4mfn, __PAGE_HYPERVISOR_RW);
     }
 
+    sl4e[shadow_l4_table_offset(LINEAR_PT_VIRT_START)] =
+        shadow_l4e_from_mfn(gl4mfn, __PAGE_HYPERVISOR_RW);
+
     unmap_domain_page(sl4e);
 }
 #endif
@@ -4406,6 +4418,11 @@ static int sh_guess_wrmap(struct vcpu *v, unsigned long vaddr, mfn_t gmfn)
 
     /* Carefully look in the shadow linear map for the l1e we expect */
 #if SHADOW_PAGING_LEVELS >= 4
+    /* Is a shadow linear map is installed in the first place? */
+    sl4p  = v->arch.paging.shadow.guest_vtable;
+    sl4p += shadow_l4_table_offset(SH_LINEAR_PT_VIRT_START);
+    if ( !(shadow_l4e_get_flags(*sl4p) & _PAGE_PRESENT) )
+        return 0;
     sl4p = sh_linear_l4_table(v) + shadow_l4_linear_offset(vaddr);
     if ( !(shadow_l4e_get_flags(*sl4p) & _PAGE_PRESENT) )
         return 0;