diff -up xf86-video-modesetting-0.8.0/src/driver.c.jx xf86-video-modesetting-0.8.0/src/driver.c
--- xf86-video-modesetting-0.8.0/src/driver.c.jx	2014-02-25 13:40:11.000000000 -0500
+++ xf86-video-modesetting-0.8.0/src/driver.c	2014-02-25 14:01:20.521092883 -0500
@@ -567,6 +567,30 @@ FreeRec(ScrnInfoPtr pScrn)
 }
 
 static Bool
+msDefaultSoftwareCursor(modesettingPtr ms)
+{
+    Bool ret = FALSE;
+    drmVersionPtr v = drmGetVersion(ms->fd);
+
+    /*
+     * This is horrific, don't hate me.  The _runtime_ fallback to software
+     * cursor doesn't seem to take on the first try; you get no cursor in
+     * gdm, but it works once you log in.  However just bombing out at
+     * init time seems to work fine.  So for drivers without hardware cursors
+     * (or for G200SE, where they're basically unusable) just default to sw.
+     */
+    if (!strcmp(v->name, "udl") ||
+	!strcmp(v->name, "mgag200") ||
+	!strcmp(v->name, "cirrus") ||
+	!strcmp(v->name, "ast"))
+	ret = TRUE;
+
+    drmFreeVersion(v);
+
+    return ret;
+}
+
+static Bool
 PreInit(ScrnInfoPtr pScrn, int flags)
 {
     modesettingPtr ms;
@@ -706,9 +730,9 @@ PreInit(ScrnInfoPtr pScrn, int flags)
     if (!xf86SetDefaultVisual(pScrn, -1))
 	return FALSE;
 
-    if (xf86ReturnOptValBool(ms->Options, OPTION_SW_CURSOR, FALSE)) {
-	ms->drmmode.sw_cursor = TRUE;
-    }
+    if (!xf86GetOptValBool(ms->Options, OPTION_SW_CURSOR,
+			   &ms->drmmode.sw_cursor))
+	ms->drmmode.sw_cursor = msDefaultSoftwareCursor(ms);
 
     if (ms->drmmode.force_24_32) {
 	    prefer_shadow = TRUE;