[PATCH] ARM: Move copy/clear user_page locking into implementation

Move the locking for copy_user_page() and clear_user_page() into
the implementations which require locking.  For simple memcpy/
memset based implementations, the locking is extra overhead which
is not necessary, and prevents preemption occuring.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>

arch/arm/mm/copypage-v4mc.S

-/*
- *  linux/arch/arm/lib/copy_page-armv4mc.S
- *
- *  Copyright (C) 1995-2001 Russell King
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
- *
- *  ASM optimised string functions
- */
-#include <linux/linkage.h>
-#include <linux/init.h>
-#include <asm/constants.h>
-
-	.text
-	.align	5
-/*
- * ARMv4 mini-dcache optimised copy_user_page
- *
- * We flush the destination cache lines just before we write the data into the
- * corresponding address.  Since the Dcache is read-allocate, this removes the
- * Dcache aliasing issue.  The writes will be forwarded to the write buffer,
- * and merged as appropriate.
- *
- * Note: We rely on all ARMv4 processors implementing the "invalidate D line"
- * instruction.  If your processor does not supply this, you have to write your
- * own copy_user_page that does the right thing.
- */
-ENTRY(v4_mc_copy_user_page)
-	stmfd	sp!, {r4, lr}			@ 2
-	mov	r4, r0
-	mov	r0, r1
-	bl	map_page_minicache
-	mov	r1, #PAGE_SZ/64			@ 1
-	ldmia	r0!, {r2, r3, ip, lr}		@ 4
-1:	mcr	p15, 0, r4, c7, c6, 1		@ 1   invalidate D line
-	stmia	r4!, {r2, r3, ip, lr}		@ 4
-	ldmia	r0!, {r2, r3, ip, lr}		@ 4+1
-	stmia	r4!, {r2, r3, ip, lr}		@ 4
-	ldmia	r0!, {r2, r3, ip, lr}		@ 4
-	mcr	p15, 0, r4, c7, c6, 1		@ 1   invalidate D line
-	stmia	r4!, {r2, r3, ip, lr}		@ 4
-	ldmia	r0!, {r2, r3, ip, lr}		@ 4
-	subs	r1, r1, #1			@ 1
-	stmia	r4!, {r2, r3, ip, lr}		@ 4
-	ldmneia	r0!, {r2, r3, ip, lr}		@ 4
-	bne	1b				@ 1
-	ldmfd	sp!, {r4, pc}			@ 3
-
-	.align	5
-/*
- * ARMv4 optimised clear_user_page
- *
- * Same story as above.
- */
-ENTRY(v4_mc_clear_user_page)
-	str	lr, [sp, #-4]!
-	mov	r1, #PAGE_SZ/64			@ 1
-	mov	r2, #0				@ 1
-	mov	r3, #0				@ 1
-	mov	ip, #0				@ 1
-	mov	lr, #0				@ 1
-1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	stmia	r0!, {r2, r3, ip, lr}		@ 4
-	subs	r1, r1, #1			@ 1
-	bne	1b				@ 1
-	ldr	pc, [sp], #4
-
-	__INITDATA
-
-	.type	v4_mc_user_fns, #object
-ENTRY(v4_mc_user_fns)
-	.long	v4_mc_clear_user_page
-	.long	v4_mc_copy_user_page
-	.size	v4_mc_user_fns, . - v4_mc_user_fns

arch/arm/mm/copypage-v4mc.c

+/*
+ *  linux/arch/arm/lib/copypage-armv4mc.S
+ *
+ *  Copyright (C) 1995-2005 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This handles the mini data cache, as found on SA11x0 and XScale
+ * processors.  When we copy a user page page, we map it in such a way
+ * that accesses to this page will not touch the main data cache, but
+ * will be cached in the mini data cache.  This prevents us thrashing
+ * the main data cache on page faults.
+ */
+#include <linux/init.h>
+#include <linux/mm.h>
+
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/tlbflush.h>
+
+/*
+ * 0xffff8000 to 0xffffffff is reserved for any ARM architecture
+ * specific hacks for copying pages efficiently.
+ */
+#define minicache_pgprot __pgprot(L_PTE_PRESENT | L_PTE_YOUNG | \
+				  L_PTE_CACHEABLE)
+
+#define TOP_PTE(x)	pte_offset_kernel(top_pmd, x)
+
+static DEFINE_SPINLOCK(minicache_lock);
+
+/*
+ * ARMv4 mini-dcache optimised copy_user_page
+ *
+ * We flush the destination cache lines just before we write the data into the
+ * corresponding address.  Since the Dcache is read-allocate, this removes the
+ * Dcache aliasing issue.  The writes will be forwarded to the write buffer,
+ * and merged as appropriate.
+ *
+ * Note: We rely on all ARMv4 processors implementing the "invalidate D line"
+ * instruction.  If your processor does not supply this, you have to write your
+ * own copy_user_page that does the right thing.
+ */
+static void __attribute__((naked))
+mc_copy_user_page(void *from, void *to)
+{
+	asm volatile(
+	"stmfd	sp!, {r4, lr}			@ 2\n\
+	mov	r4, %2				@ 1\n\
+	ldmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+1:	mcr	p15, 0, %1, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	%1!, {r2, r3, ip, lr}		@ 4\n\
+	ldmia	%0!, {r2, r3, ip, lr}		@ 4+1\n\
+	stmia	%1!, {r2, r3, ip, lr}		@ 4\n\
+	ldmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	mcr	p15, 0, %1, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	%1!, {r2, r3, ip, lr}		@ 4\n\
+	ldmia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	subs	r4, r4, #1			@ 1\n\
+	stmia	%1!, {r2, r3, ip, lr}		@ 4\n\
+	ldmneia	%0!, {r2, r3, ip, lr}		@ 4\n\
+	bne	1b				@ 1\n\
+	ldmfd	sp!, {r4, pc}			@ 3"
+	:
+	: "r" (from), "r" (to), "I" (PAGE_SIZE / 64));
+}
+
+void v4_mc_copy_user_page(void *kto, const void *kfrom, unsigned long vaddr)
+{
+	spin_lock(&minicache_lock);
+
+	set_pte(TOP_PTE(COPYPAGE_MINICACHE), pfn_pte(__pa(kfrom) >> PAGE_SHIFT, minicache_pgprot));
+	flush_tlb_kernel_page(COPYPAGE_MINICACHE);
+
+	mc_copy_user_page((void *)COPYPAGE_MINICACHE, kto);
+
+	spin_unlock(&minicache_lock);
+}
+
+/*
+ * ARMv4 optimised clear_user_page
+ */
+void __attribute__((naked))
+v4_mc_clear_user_page(void *kaddr, unsigned long vaddr)
+{
+	asm volatile(
+	"str	lr, [sp, #-4]!\n\
+	mov	r1, %0				@ 1\n\
+	mov	r2, #0				@ 1\n\
+	mov	r3, #0				@ 1\n\
+	mov	ip, #0				@ 1\n\
+	mov	lr, #0				@ 1\n\
+1:	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+	mcr	p15, 0, r0, c7, c6, 1		@ 1   invalidate D line\n\
+	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+	stmia	r0!, {r2, r3, ip, lr}		@ 4\n\
+	subs	r1, r1, #1			@ 1\n\
+	bne	1b				@ 1\n\
+	ldr	pc, [sp], #4"
+	:
+	: "I" (PAGE_SIZE / 64));
+}
+
+struct cpu_user_fns v4_mc_user_fns __initdata = {
+	.cpu_clear_user_page	= v4_mc_clear_user_page, 
+	.cpu_copy_user_page	= v4_mc_copy_user_page,
+};

include/asm-arm/page.h

@@ -114,19 +114,8 @@ extern void __cpu_copy_user_page(void *to, const void *from,
 				 unsigned long user);
 #endif
 
-#define clear_user_page(addr,vaddr,pg)			\
-	do {						\
-		preempt_disable();			\
-		__cpu_clear_user_page(addr, vaddr);	\
-		preempt_enable();			\
-	} while (0)
-
-#define copy_user_page(to,from,vaddr,pg)		\
-	do {						\
-		preempt_disable();			\
-		__cpu_copy_user_page(to, from, vaddr);	\
-		preempt_enable();			\
-	} while (0)
+#define clear_user_page(addr,vaddr,pg)	 __cpu_clear_user_page(addr, vaddr)
+#define copy_user_page(to,from,vaddr,pg) __cpu_copy_user_page(to, from, vaddr)
 
 #define clear_page(page)	memzero((void *)(page), PAGE_SIZE)
 extern void copy_page(void *to, const void *from);