sh: add kexec jump support

Add kexec jump support to the SuperH architecture.

Similar to the x86 implementation, with the following
exceptions:

- Instead of separating the assembly code flow into
two parts for regular kexec and kexec jump we use a
single code path. In the assembly snippet regular
kexec is just kexec jump that never comes back.

- Instead of using a swap page when moving data between
pages the page copy assembly routine has been modified
to exchange the data between the pages using registers.

- We walk the page list twice in machine_kexec() to
do and undo physical to virtual address conversion.

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

arch/sh/kernel/relocate_kernel.S

@@ -4,6 +4,8 @@
  *
  * LANDISK/sh4 is supported. Maybe, SH archtecture works well.
  *
+ * 2009-03-18 Magnus Damm - Added Kexec Jump support
+ *
  * This source code is licensed under the GNU General Public License,
  * Version 2.  See the file COPYING for more details.
  */
@@ -17,14 +19,135 @@ relocate_new_kernel:
 	/* r5 = reboot_code_buffer */
 	/* r6 = start_address      */
 
-	mov.l	10f,r8	  /* PAGE_SIZE */
+	mov.l	10f, r0		/* PAGE_SIZE */
+	add	r5, r0		/* setup new stack at end of control page */
 
-	/*  stack setting */
-	add	r8,r5
-	mov	r5,r15
+	/* save r15->r8 to new stack */
+	mov.l	r15, @-r0
+	mov	r0, r15
+	mov.l	r14, @-r15
+	mov.l	r13, @-r15
+	mov.l	r12, @-r15
+	mov.l	r11, @-r15
+	mov.l	r10, @-r15
+	mov.l	r9, @-r15
+	mov.l	r8, @-r15
 
+	/* save other random registers */
+	sts.l	macl, @-r15
+	sts.l	mach, @-r15
+	stc.l	gbr, @-r15
+	stc.l	ssr, @-r15
+	stc.l	sr, @-r15
+	sts.l	pr, @-r15
+	stc.l	spc, @-r15
+
+	/* switch to bank1 and save r7->r0 */
+	mov.l	12f, r9
+	stc	sr, r8
+	or	r9, r8
+	ldc	r8, sr
+	mov.l	r7, @-r15
+	mov.l	r6, @-r15
+	mov.l	r5, @-r15
+	mov.l	r4, @-r15
+	mov.l	r3, @-r15
+	mov.l	r2, @-r15
+	mov.l	r1, @-r15
+	mov.l	r0, @-r15
+
+	/* switch to bank0 and save r7->r0 */
+	mov.l	12f, r9
+	not	r9, r9
+	stc	sr, r8
+	and	r9, r8
+	ldc	r8, sr
+	mov.l	r7, @-r15
+	mov.l	r6, @-r15
+	mov.l	r5, @-r15
+	mov.l	r4, @-r15
+	mov.l	r3, @-r15
+	mov.l	r2, @-r15
+	mov.l	r1, @-r15
+	mov.l	r0, @-r15
+
+	mov.l	r4, @-r15	/* save indirection page again */
+
+	bsr	swap_pages	/* swap pages before jumping to new kernel */
+	 nop
+
+	mova	11f, r0
+	mov.l	r15, @r0	/* save pointer to stack */
+
+	jsr	@r6		/* hand over control to new kernel */
+	 nop
+
+	mov.l	11f, r15	/* get pointer to stack */
+	mov.l	@r15+, r4	/* restore r4 to get indirection page */
+
+	bsr	swap_pages	/* swap pages back to previous state */
+	 nop
+
+	/* make sure bank0 is active and restore r0->r7 */
+	mov.l	12f, r9
+	not	r9, r9
+	stc	sr, r8
+	and	r9, r8
+	ldc	r8, sr
+	mov.l	@r15+, r0
+	mov.l	@r15+, r1
+	mov.l	@r15+, r2
+	mov.l	@r15+, r3
+	mov.l	@r15+, r4
+	mov.l	@r15+, r5
+	mov.l	@r15+, r6
+	mov.l	@r15+, r7
+
+	/* switch to bank1 and restore r0->r7 */
+	mov.l	12f, r9
+	stc	sr, r8
+	or	r9, r8
+	ldc	r8, sr
+	mov.l	@r15+, r0
+	mov.l	@r15+, r1
+	mov.l	@r15+, r2
+	mov.l	@r15+, r3
+	mov.l	@r15+, r4
+	mov.l	@r15+, r5
+	mov.l	@r15+, r6
+	mov.l	@r15+, r7
+
+	/* switch back to bank0 */
+	mov.l	12f, r9
+	not	r9, r9
+	stc	sr, r8
+	and	r9, r8
+	ldc	r8, sr
+
+	/* restore other random registers */
+	ldc.l	@r15+, spc
+	lds.l	@r15+, pr
+	ldc.l	@r15+, sr
+	ldc.l	@r15+, ssr
+	ldc.l	@r15+, gbr
+	lds.l	@r15+, mach
+	lds.l	@r15+, macl
+
+	/* restore r8->r15 */
+	mov.l	@r15+, r8
+	mov.l	@r15+, r9
+	mov.l	@r15+, r10
+	mov.l	@r15+, r11
+	mov.l	@r15+, r12
+	mov.l	@r15+, r13
+	mov.l	@r15+, r14
+	mov.l	@r15+, r15
+	rts
+	 nop
+
+swap_pages:
 	bra	1f
-	mov	r4,r0	  /* cmd = indirection_page */
+	 mov	r4,r0	  /* cmd = indirection_page */
 0:
 	mov.l	@r4+,r0	  /* cmd = *ind++ */
 
@@ -37,52 +160,70 @@ relocate_new_kernel:
 	tst	#1,r0
 	bt	2f
 	bra	0b
-	mov	r2,r5
+	 mov	r2,r5
 
 2:	/* else if(cmd & IND_INDIRECTION) ind = addr  */
 	tst	#2,r0
 	bt	3f
 	bra	0b
-	mov	r2,r4
+	 mov	r2,r4
 
-3:	/* else if(cmd & IND_DONE) goto 6  */
+3:	/* else if(cmd & IND_DONE) return */
 	tst	#4,r0
 	bt	4f
-	bra	6f
-	nop
+	rts
+	 nop
 
 4:	/* else if(cmd & IND_SOURCE) memcpy(dst,addr,PAGE_SIZE) */
 	tst	#8,r0
 	bt	0b
 
-	mov	r8,r3
+	mov.l	10f,r3	  /* PAGE_SIZE */
 	shlr2	r3
 	shlr2	r3
 5:
 	dt	r3
-	mov.l	@r2+,r1   /*  16n+0 */
-	mov.l	r1,@r5
-	add	#4,r5
-	mov.l	@r2+,r1	  /*  16n+4 */
-	mov.l	r1,@r5
-	add	#4,r5
-	mov.l	@r2+,r1   /*  16n+8 */
-	mov.l	r1,@r5
-	add	#4,r5
-	mov.l	@r2+,r1   /*  16n+12 */
-	mov.l	r1,@r5
-	add	#4,r5
+
+	/* regular kexec just overwrites the destination page
+	 * with the contents of the source page.
+	 * for the kexec jump case we need to swap the contents
+	 * of the pages.
+	 * to keep it simple swap the contents for both cases.
+	 */
+	mov.l	@(0, r2), r8
+	mov.l	@(0, r5), r1
+	mov.l	r8, @(0, r5)
+	mov.l	r1, @(0, r2)
+
+	mov.l	@(4, r2), r8
+	mov.l	@(4, r5), r1
+	mov.l	r8, @(4, r5)
+	mov.l	r1, @(4, r2)
+
+	mov.l	@(8, r2), r8
+	mov.l	@(8, r5), r1
+	mov.l	r8, @(8, r5)
+	mov.l	r1, @(8, r2)
+
+	mov.l	@(12, r2), r8
+	mov.l	@(12, r5), r1
+	mov.l	r8, @(12, r5)
+	mov.l	r1, @(12, r2)
+
+	add	#16,r5
+	add	#16,r2
 	bf	5b
 
 	bra	0b
-	nop
-6:
-	jmp @r6
-	nop
+	 nop
 
 	.align 2
 10:
 	.long	PAGE_SIZE
+11:
+	.long	0
+12:
+	.long	0x20000000 ! RB=1
 
 relocate_new_kernel_end: