sh: shared register saving code for sh3/sh4/sh4a

This patch reworks the sh3/sh4/sh4a register saving code in
the following ways:
 - break out prepare_stack_save_dsp() from handle_exception()
 - break out save_regs() from handle_exception()
 - the register saving order is unchanged
 - align new functions to fit in cache lines
 - separate exception code from interrupt code
 - keep main code flow in a single cache line per exception vector
 - use bsr/rts for regular functions (save pr first)
 - keep data in one shared cache line (exception_data)
 - document the functions
 - tie in the hp6xx code

Signed-off-by: Magnus Damm <damm@igel.co.jp>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

arch/sh/boards/mach-hp6xx/pm_wakeup.S

@@ -10,47 +10,32 @@
 #include <linux/linkage.h>
 #include <cpu/mmu_context.h>
 
-#define k0	r0
-#define k1	r1
-#define k2	r2
-#define k3	r3
-#define k4	r4
-
 /*
  * Kernel mode register usage:
  *	k0	scratch
  *	k1	scratch
- *	k2	scratch (Exception code)
- *	k3	scratch (Return address)
- *	k4	scratch
- *	k5	reserved
- *	k6	Global Interrupt Mask (0--15 << 4)
- *	k7	CURRENT_THREAD_INFO (pointer to current thread info)
+ * For more details, please have a look at entry.S
  */
 
+#define k0	r0
+#define k1	r1
+
 ENTRY(wakeup_start)
 ! clear STBY bit
-	mov	#-126, k2
+	mov	#-126, k1
    	and	#127, k0
-	mov.b	k0, @k2
+	mov.b	k0, @k1
 ! enable refresh
 	mov.l	5f, k1
 	mov.w	6f, k0
   	mov.w	k0, @k1
 ! jump to handler
-	mov.l	2f, k2
-	mov.l	3f, k3
-	mov.l	@k2, k2
-
 	mov.l	4f, k1
 	jmp	@k1
-	nop
+	 nop
 
 	.align	2
-1:	.long	EXPEVT
-2:	.long	INTEVT
-3:	.long	ret_from_irq
-4:	.long	handle_exception
+4:	.long	handle_interrupt
 5:	.long	0xffffff68
 6:	.word	0x0524