appliance: Use -cpu max.

QEMU has a newish feature (from about 2017 / qemu 2.9) called -cpu max
which is supposed to select the best CPU, ideal for libguestfs.

After this change, on x86-64:

               KVM                          TCG

Direct         -cpu max                     -cpu max
(non-libvirt)

Libvirt   <cpu mode="host-passthrough">     <cpu mode="host-model">
            <model fallback="allow"/>         <model fallback="allow"/>
          </cpu>                            </cpu>

Thanks: Daniel Berrangé

lib/appliance-cpu.c

@@ -38,6 +38,11 @@
  *
  * The literal string C<"host"> means use C<-cpu host>.
  *
+ * =item C<"max">
+ *
+ * The literal string C<"max"> means use C<-cpu max> (the best
+ * possible).  This requires awkward translation for libvirt.
+ *
  * =item some string
  *
  * Some string such as C<"cortex-a57"> means use C<-cpu cortex-a57>.
@@ -80,14 +85,9 @@ guestfs_int_get_cpu_model (int kvm)
   /* See discussion in https://bugzilla.redhat.com/show_bug.cgi?id=1605071 */
   return NULL;
 #else
-  /* On most architectures, it is faster to pass the CPU host model to
-   * the appliance, allowing maximum speed for things like checksums
-   * and encryption.  Only do this with KVM.  It is broken in subtle
-   * ways on TCG, and fairly pointless when you're emulating anyway.
+  /* On most architectures we can use "max" to get the best possible CPU.
+   * For recent qemu this should work even on TCG.
    */
-  if (kvm)
-    return "host";
-  else
-    return NULL;
+  return "max";
 #endif
 }

lib/launch-libvirt.c

@@ -1169,6 +1169,15 @@ construct_libvirt_xml_cpu (guestfs_h *g,
           attribute ("fallback", "allow");
         } end_element ();
       }
+      else if (STREQ (cpu_model, "max")) {
+        if (params->data->is_kvm)
+          attribute ("mode", "host-passthrough");
+        else
+          attribute ("mode", "host-model");
+        start_element ("model") {
+          attribute ("fallback", "allow");
+        } end_element ();
+      }
       else
         single_element ("model", cpu_model);
     } end_element ();