fw_cfg: unbreak migration compatibility for 2.4 and earlier machines

When I reviewed Marc's fw_cfg DMA patches, I completely missed that the
way we set dma_enabled would break migration.

Gerd explained the right way (see reference below): dma_enabled should be
set to true by default, and only true->false transitions should be
possible:

- when the user requests that with

    -global fw_cfg_mem.dma_enabled=off

  or

   -global fw_cfg_io.dma_enabled=off

  as appropriate for the platform,

- when HW_COMPAT_2_4 dictates it,

- when board code initializes fw_cfg without requesting DMA support.

Cc: Marc Marí <markmb@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Alexandre DERUMIER <aderumier@odiso.com>
Cc: qemu-stable@nongnu.org
Ref: http://thread.gmane.org/gmane.comp.emulators.qemu/390272/focus=391042
Ref: https://bugs.launchpad.net/qemu/+bug/1536487
Suggested-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-id: 1455823860-22268-1-git-send-email-lersek@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

include/hw/compat.h

@@ -42,6 +42,14 @@
         .driver   = "virtio-pci",\
         .property = "migrate-extra",\
         .value    = "off",\
+    },{\
+        .driver   = "fw_cfg_mem",\
+        .property = "dma_enabled",\
+        .value    = "off",\
+    },{\
+        .driver   = "fw_cfg_io",\
+        .property = "dma_enabled",\
+        .value    = "off",\
     },
 
 #define HW_COMPAT_2_3 \