sh: Fixup various PAGE_SIZE == 4096 assumptions.

There were a number of places that made evil PAGE_SIZE == 4k
assumptions that ended up breaking when trying to play with
8k and 64k page sizes, this fixes those up.

The most significant change is the way we load THREAD_SIZE,
previously this was done via:

	mov	#(THREAD_SIZE >> 8), reg
	shll8	reg

to avoid a memory access and allow the immediate load. With
a 64k PAGE_SIZE, we're out of range for the immediate load
size without resorting to special instructions available in
later ISAs (movi20s and so on). The "workaround" for this is
to bump up the shift to 10 and insert a shll2, which gives a
bit more flexibility while still being much cheaper than a
memory access.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>

arch/sh/kernel/cpu/sh3/entry.S

@@ -335,7 +335,7 @@ general_exception:
 /* This code makes some assumptions to improve performance.
  * Make sure they are stil true. */
 #if PTRS_PER_PGD != PTRS_PER_PTE
-#error PDG and PTE sizes don't match
+#error PGD and PTE sizes don't match
 #endif
 
 /* gas doesn't flag impossible values for mov #immediate as an error */
@@ -547,8 +547,9 @@ ENTRY(handle_exception)
 	bt/s	1f		! It's a kernel to kernel transition.
 	 mov	r15, k0		! save original stack to k0
 	/* User space to kernel */
-	mov	#(THREAD_SIZE >> 8), k1
+	mov	#(THREAD_SIZE >> 10), k1
 	shll8	k1		! k1 := THREAD_SIZE
+	shll2	k1
 	add	current, k1
 	mov	k1, r15		! change to kernel stack
 	!

arch/sh/kernel/head.S

@@ -33,7 +33,7 @@ ENTRY(empty_zero_page)
 	.long	0x00360000	/* INITRD_START */
 	.long	0x000a0000	/* INITRD_SIZE */
 	.long	0
-	.balign 4096,0,4096
+	.balign PAGE_SIZE,0,PAGE_SIZE
 
 	.text	
 /*
@@ -70,8 +70,9 @@ ENTRY(_stext)
 	!
 	mov.l	2f, r0
 	mov	r0, r15		! Set initial r15 (stack pointer)
-	mov	#(THREAD_SIZE >> 8), r1
+	mov	#(THREAD_SIZE >> 10), r1
 	shll8	r1		! r1 = THREAD_SIZE
+	shll2	r1
 	sub	r1, r0		!
 #ifdef CONFIG_CPU_HAS_SR_RB
 	ldc	r0, r7_bank	! ... and initial thread_info

arch/sh/kernel/relocate_kernel.S

@@ -7,11 +7,9 @@
  * This source code is licensed under the GNU General Public License,
  * Version 2.  See the file COPYING for more details.
  */
-
 #include <linux/linkage.h>
-
+#include <asm/addrspace.h>
-#define PAGE_SIZE      4096 /* must be same value as in <asm/page.h> */
+#include <asm/page.h>
-
 
 		.globl relocate_new_kernel
 relocate_new_kernel:
@@ -20,8 +18,8 @@ relocate_new_kernel:
 	/* r6 = start_address      */
 	/* r7 = vbr_reg            */
 
-	mov.l	10f,r8	  /* 4096 */
+	mov.l	10f,r8	  /* PAGE_SIZE */
-	mov.l	11f,r9    /* 0xa0000000 */
+	mov.l	11f,r9    /* P2SEG */
 
 	/*  stack setting */
 	add	r8,r5
@@ -32,7 +30,7 @@ relocate_new_kernel:
 0:
 	mov.l	@r4+,r0	  /* cmd = *ind++ */
 
-1:	/* addr = (cmd | 0xa0000000) & 0xfffffff0 */
+1:	/* addr = (cmd | P2SEG) & 0xfffffff0 */
 	mov	r0,r2
 	or	r9,r2
 	mov	#-16,r1
@@ -92,7 +90,7 @@ relocate_new_kernel:
 10:
 	.long	PAGE_SIZE
 11:
-	.long	0xa0000000
+	.long	P2SEG
 
 relocate_new_kernel_end:
 

arch/sh/mm/cache-sh4.c

@@ -225,7 +225,7 @@ static inline void flush_cache_4096(unsigned long start,
 	 */
 	if ((cpu_data->flags & CPU_HAS_P2_FLUSH_BUG) ||
 	    (start < CACHE_OC_ADDRESS_ARRAY))
-	    	exec_offset = 0x20000000;
+		exec_offset = 0x20000000;
 
 	local_irq_save(flags);
 	__flush_cache_4096(start | SH_CACHE_ASSOC,
@@ -246,7 +246,7 @@ void flush_dcache_page(struct page *page)
 
 		/* Loop all the D-cache */
 		n = cpu_data->dcache.n_aliases;
-		for (i = 0; i < n; i++, addr += PAGE_SIZE)
+		for (i = 0; i < n; i++, addr += 4096)
 			flush_cache_4096(addr, phys);
 	}
 

arch/sh/mm/clear_page.S

@@ -1,12 +1,12 @@
-/* $Id: clear_page.S,v 1.13 2003/08/25 17:03:10 lethal Exp $
+/*
- *
  * __clear_user_page, __clear_user, clear_page implementation of SuperH
  *
  * Copyright (C) 2001  Kaz Kojima
  * Copyright (C) 2001, 2002  Niibe Yutaka
- *
+ * Copyright (C) 2006  Paul Mundt
  */
 #include <linux/linkage.h>
+#include <asm/page.h>
 
 /*
  * clear_page_slow
@@ -18,11 +18,11 @@
 /*
  * r0 --- scratch
  * r4 --- to
- * r5 --- to + 4096
+ * r5 --- to + PAGE_SIZE
  */
 ENTRY(clear_page_slow)
 	mov	r4,r5
-	mov.w	.Llimit,r0
+	mov.l	.Llimit,r0
 	add	r0,r5
 	mov	#0,r0
 	!
@@ -50,7 +50,7 @@ ENTRY(clear_page_slow)
 	!
 	rts
 	 nop
-.Llimit:	.word	(4096-28)
+.Llimit:	.long	(PAGE_SIZE-28)
 
 ENTRY(__clear_user)
 	!
@@ -164,10 +164,10 @@ ENTRY(__clear_user)
  * r0 --- scratch 
  * r4 --- to
  * r5 --- orig_to
- * r6 --- to + 4096
+ * r6 --- to + PAGE_SIZE
  */
 ENTRY(__clear_user_page)
-	mov.w	.L4096,r0
+	mov.l	.Lpsz,r0
 	mov	r4,r6
 	add	r0,r6
 	mov	#0,r0
@@ -191,7 +191,7 @@ ENTRY(__clear_user_page)
 	!
 	rts
 	 nop
-.L4096:	.word	4096
+.Lpsz:	.long	PAGE_SIZE
 
 #endif
 

arch/sh/mm/copy_page.S

@@ -1,12 +1,12 @@
-/* $Id: copy_page.S,v 1.8 2003/08/25 17:03:10 lethal Exp $
+/*
- *
  * copy_page, __copy_user_page, __copy_user implementation of SuperH
  *
  * Copyright (C) 2001  Niibe Yutaka & Kaz Kojima
  * Copyright (C) 2002  Toshinobu Sugioka
- *
+ * Copyright (C) 2006  Paul Mundt
  */
 #include <linux/linkage.h>
+#include <asm/page.h>
 
 /*
  * copy_page_slow
@@ -18,7 +18,7 @@
 
 /*
  * r0, r1, r2, r3, r4, r5, r6, r7 --- scratch 
- * r8 --- from + 4096
+ * r8 --- from + PAGE_SIZE
  * r9 --- not used
  * r10 --- to
  * r11 --- from
@@ -30,7 +30,7 @@ ENTRY(copy_page_slow)
 	mov	r4,r10
 	mov	r5,r11
 	mov	r5,r8
-	mov.w	.L4096,r0
+	mov.l	.Lpsz,r0
 	add	r0,r8
 	!
 1:	mov.l	@r11+,r0
@@ -80,7 +80,7 @@ ENTRY(copy_page_slow)
 
 /*
  * r0, r1, r2, r3, r4, r5, r6, r7 --- scratch 
- * r8 --- from + 4096
+ * r8 --- from + PAGE_SIZE
  * r9 --- orig_to
  * r10 --- to
  * r11 --- from
@@ -94,7 +94,7 @@ ENTRY(__copy_user_page)
 	mov	r5,r11
 	mov	r6,r9
 	mov	r5,r8
-	mov.w	.L4096,r0
+	mov.l	.Lpsz,r0
 	add	r0,r8
 	!
 1:	ocbi	@r9
@@ -129,7 +129,7 @@ ENTRY(__copy_user_page)
 	rts
 	 nop
 #endif
-.L4096:	.word	4096
+.Lpsz:	.long	PAGE_SIZE
 /*
  * __kernel_size_t __copy_user(void *to, const void *from, __kernel_size_t n);
  * Return the number of bytes NOT copied

arch/sh/mm/init.c

@@ -217,7 +217,6 @@ static struct kcore_list kcore_mem, kcore_vmalloc;
 
 void __init mem_init(void)
 {
-	extern unsigned long empty_zero_page[1024];
 	int codesize, reservedpages, datasize, initsize;
 	int tmp;
 	extern unsigned long memory_start;

arch/sh/mm/pg-dma.c

@@ -39,8 +39,6 @@ static void copy_page_dma(void *to, void *from)
 
 static void clear_page_dma(void *to)
 {
-	extern unsigned long empty_zero_page[1024];
-
 	/*
 	 * We get invoked quite early on, if the DMAC hasn't been initialized
 	 * yet, fall back on the slow manual implementation.

include/asm-sh/entry-macros.S

@@ -23,8 +23,9 @@
 #ifdef CONFIG_HAS_SR_RB
 	stc	r7_bank, \ti
 #else
-	mov	#((THREAD_SIZE - 1)>> 8) ^ 0xff, \tmp
+	mov	#((THREAD_SIZE - 1) >> 10) ^ 0xff, \tmp
 	shll8	\tmp
+	shll2	\tmp
 	mov	r15, \ti
 	and	\tmp, \ti
 #endif	

include/asm-sh/pgtable.h

@@ -47,8 +47,8 @@ extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
 #define PGDIR_MASK	(~(PGDIR_SIZE-1))
 
 /* Entries per level */
-#define PTRS_PER_PTE	(1 << PTE_BITS)
+#define PTRS_PER_PTE	(PAGE_SIZE / 4)
-#define PTRS_PER_PGD	(1 << PGDIR_BITS)
+#define PTRS_PER_PGD	(PAGE_SIZE / 4)
 
 #define USER_PTRS_PER_PGD	(TASK_SIZE/PGDIR_SIZE)
 #define FIRST_USER_ADDRESS	0
@@ -57,7 +57,7 @@ extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
 
 /*
  * First 1MB map is used by fixed purpose.
- * Currently only 4-enty (16kB) is used (see arch/sh/mm/cache.c)
+ * Currently only 4-entry (16kB) is used (see arch/sh/mm/cache.c)
  */
 #define VMALLOC_START	(P3SEG+0x00100000)
 #define VMALLOC_END	(FIXADDR_START-2*PAGE_SIZE)

include/asm-sh/thread_info.h

@@ -90,13 +90,7 @@ static inline struct thread_info *current_thread_info(void)
 #endif
 #define free_thread_info(ti)	kfree(ti)
 
-#else /* !__ASSEMBLY__ */
+#endif /* __ASSEMBLY__ */
-
-/* how to get the thread information struct from ASM */
-#define GET_THREAD_INFO(reg) \
-	stc     r7_bank, reg
-
-#endif
 
 /*
  * thread information flags