ia64: Fix __delay() even more: it's just not safe to inline the delay-

	loop because, e.g., the compiler might decide to unroll the
	loop when passed a constant, etc.

arch/ia64/kernel/head.S

@@ -797,6 +797,25 @@ GLOBAL_ENTRY(ia64_switch_mode_virt)
 	br.ret.sptk.many rp
 END(ia64_switch_mode_virt)
 
+GLOBAL_ENTRY(ia64_delay_loop)
+	.prologue
+{	nop 0			// work around GAS unwind info generation bug...
+	.save ar.lc,r2
+	mov r2=ar.lc
+	.body
+	;;
+	mov ar.lc=r32
+}
+	;;
+	// force loop to be 32-byte aligned (GAS bug means we cannot use .align
+	// inside function body without corrupting unwind info).
+{	nop 0 }
+1:	br.cloop.sptk.few 1b
+	;;
+	mov ar.lc=r2
+	br.ret.sptk.many rp
+END(ia64_delay_loop)
+
 #ifdef CONFIG_IA64_BRL_EMU
 
 /*

include/asm-ia64/delay.h

@@ -67,14 +67,15 @@ ia64_get_itc (void)
 	return result;
 }
 
+extern void ia64_delay_loop (unsigned long loops);
+
 static __inline__ void
 __delay (unsigned long loops)
 {
-	if (loops < 1)
+	if (unlikely(loops < 1))
 		return;
 
-	while (loops--)
-		barrier();
+	ia64_delay_loop (loops - 1);
 }
 
 static __inline__ void