insert a magical second memory barrier prior to calling draina() in

badaddr_read().  This fixes 'machine check in pal mode' halts on
ev5 2100As.

MFC candidate -- after spending 6 hours tracking this down, I checked and
discovered that it has been in NetBSD for over a year, so it should be safe
for MFC into 4.3-RELEASE

sys/alpha/alpha/interrupt.c

@@ -283,6 +283,7 @@ badaddr_read(addr, size, rptr)
 		panic("badaddr: invalid size (%ld)\n", size);
 	}
 	alpha_mb();
+	alpha_mb(); /* magic for ev5 2100A  & maybe more */
 
 	/* Make sure we took the machine check, if we caused one. */
 	alpha_pal_draina();