s390/string: provide asm lib functions for memcpy and memcmp

Our memcpy and memcmp variants were implemented by calling the corresponding
gcc builtin variants.
However gcc is free to replace a call to __builtin_memcmp with a call to memcmp
which, when called, will result in an endless recursion within memcmp.
So let's provide asm variants and also fix the variants that are used for
uncompressing the kernel image.
In addition remove all other occurences of builtin function calls.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

7 files changed, 213 insertions, 81 deletions

diff --git a/arch/s390/boot/compressed/misc.c b/arch/s390/boot/compressed/misc.c
index 465eca7..c4c6a1c 100644
--- a/arch/s390/boot/compressed/misc.c
+++ b/arch/s390/boot/compressed/misc.c
@@ -71,34 +71,37 @@ void *memset(void *s, int c, size_t n)
 {
 	char *xs;
 
-	if (c == 0)
-		return __builtin_memset(s, 0, n);
-
-	xs = (char *) s;
-	if (n > 0)
-		do {
-			*xs++ = c;
-		} while (--n > 0);
+	xs = s;
+	while (n--)
+		*xs++ = c;
 	return s;
 }
 
-void *memcpy(void *__dest, __const void *__src, size_t __n)
+void *memcpy(void *dest, const void *src, size_t n)
 {
-	return __builtin_memcpy(__dest, __src, __n);
+	const char *s = src;
+	char *d = dest;
+
+	while (n--)
+		*d++ = *s++;
+	return dest;
 }
 
-void *memmove(void *__dest, __const void *__src, size_t __n)
+void *memmove(void *dest, const void *src, size_t n)
 {
-	char *d;
-	const char *s;
-
-	if (__dest <= __src)
-		return __builtin_memcpy(__dest, __src, __n);
-	d = __dest + __n;
-	s = __src + __n;
-	while (__n--)
-		*--d = *--s;
-	return __dest;
+	const char *s = src;
+	char *d = dest;
+
+	if (d <= s) {
+		while (n--)
+			*d++ = *s++;
+	} else {
+		d += n;
+		s += n;
+		while (n--)
+			*--d = *--s;
+	}
+	return dest;
 }
 
 static void error(char *x)
diff --git a/arch/s390/include/asm/string.h b/arch/s390/include/asm/string.h
index 1bd1352..7e2dcd7 100644
--- a/arch/s390/include/asm/string.h
+++ b/arch/s390/include/asm/string.h
@@ -96,7 +96,6 @@ static inline char *strcat(char *dst, const char *src)
 
 static inline char *strcpy(char *dst, const char *src)
 {
-#if __GNUC__ < 4
 	register int r0 asm("0") = 0;
 	char *ret = dst;
 
@@ -106,14 +105,10 @@ static inline char *strcpy(char *dst, const char *src)
 		: "+&a" (dst), "+&a" (src) : "d" (r0)
 		: "cc", "memory");
 	return ret;
-#else
-	return __builtin_strcpy(dst, src);
-#endif
 }
 
 static inline size_t strlen(const char *s)
 {
-#if __GNUC__ < 4
 	register unsigned long r0 asm("0") = 0;
 	const char *tmp = s;
 
@@ -122,9 +117,6 @@ static inline size_t strlen(const char *s)
 		"	jo	0b"
 		: "+d" (r0), "+a" (tmp) :  : "cc");
 	return r0 - (unsigned long) s;
-#else
-	return __builtin_strlen(s);
-#endif
 }
 
 static inline size_t strnlen(const char * s, size_t n)
diff --git a/arch/s390/kernel/s390_ksyms.c b/arch/s390/kernel/s390_ksyms.c
index 57b5366..9bdbcef 100644
--- a/arch/s390/kernel/s390_ksyms.c
+++ b/arch/s390/kernel/s390_ksyms.c
@@ -8,3 +8,5 @@ EXPORT_SYMBOL(_mcount);
 #if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
 EXPORT_SYMBOL(sie64a);
 #endif
+EXPORT_SYMBOL(memcpy);
+EXPORT_SYMBOL(memset);
diff --git a/arch/s390/lib/Makefile b/arch/s390/lib/Makefile
index 761ab8b..6ab0d0b 100644
--- a/arch/s390/lib/Makefile
+++ b/arch/s390/lib/Makefile
@@ -4,6 +4,7 @@
 
 lib-y += delay.o string.o uaccess_std.o uaccess_pt.o
 obj-y += usercopy.o
-obj-$(CONFIG_32BIT) += div64.o qrnnd.o ucmpdi2.o
+obj-$(CONFIG_32BIT) += div64.o qrnnd.o ucmpdi2.o mem32.o
+obj-$(CONFIG_64BIT) += mem64.o
 lib-$(CONFIG_64BIT) += uaccess_mvcos.o
 lib-$(CONFIG_SMP) += spinlock.o
diff --git a/arch/s390/lib/mem32.S b/arch/s390/lib/mem32.S
new file mode 100644
index 0000000..14ca924
--- /dev/null
+++ b/arch/s390/lib/mem32.S
@@ -0,0 +1,92 @@
+/*
+ * String handling functions.
+ *
+ * Copyright IBM Corp. 2012
+ */
+
+#include <linux/linkage.h>
+
+/*
+ * memset implementation
+ *
+ * This code corresponds to the C construct below. We do distinguish
+ * between clearing (c == 0) and setting a memory array (c != 0) simply
+ * because nearly all memset invocations in the kernel clear memory and
+ * the xc instruction is preferred in such cases.
+ *
+ * void *memset(void *s, int c, size_t n)
+ * {
+ *	if (likely(c == 0))
+ *		return __builtin_memset(s, 0, n);
+ *	return __builtin_memset(s, c, n);
+ * }
+ */
+ENTRY(memset)
+	basr	%r5,%r0
+.Lmemset_base:
+	ltr	%r4,%r4
+	bzr	%r14
+	ltr	%r3,%r3
+	jnz	.Lmemset_fill
+	ahi	%r4,-1
+	lr	%r3,%r4
+	srl	%r3,8
+	ltr	%r3,%r3
+	lr	%r1,%r2
+	je	.Lmemset_clear_rest
+.Lmemset_clear_loop:
+	xc	0(256,%r1),0(%r1)
+	la	%r1,256(%r1)
+	brct	%r3,.Lmemset_clear_loop
+.Lmemset_clear_rest:
+	ex	%r4,.Lmemset_xc-.Lmemset_base(%r5)
+	br	%r14
+.Lmemset_fill:
+	stc	%r3,0(%r2)
+	chi	%r4,1
+	lr	%r1,%r2
+	ber	%r14
+	ahi	%r4,-2
+	lr	%r3,%r4
+	srl	%r3,8
+	ltr	%r3,%r3
+	je	.Lmemset_fill_rest
+.Lmemset_fill_loop:
+	mvc	1(256,%r1),0(%r1)
+	la	%r1,256(%r1)
+	brct	%r3,.Lmemset_fill_loop
+.Lmemset_fill_rest:
+	ex	%r4,.Lmemset_mvc-.Lmemset_base(%r5)
+	br	%r14
+.Lmemset_xc:
+	xc	0(1,%r1),0(%r1)
+.Lmemset_mvc:
+	mvc	1(1,%r1),0(%r1)
+
+/*
+ * memcpy implementation
+ *
+ * void *memcpy(void *dest, const void *src, size_t n)
+ */
+ENTRY(memcpy)
+	basr	%r5,%r0
+.Lmemcpy_base:
+	ltr	%r4,%r4
+	bzr	%r14
+	ahi	%r4,-1
+	lr	%r0,%r4
+	srl	%r0,8
+	ltr	%r0,%r0
+	lr	%r1,%r2
+	jnz	.Lmemcpy_loop
+.Lmemcpy_rest:
+	ex	%r4,.Lmemcpy_mvc-.Lmemcpy_base(%r5)
+	br	%r14
+.Lmemcpy_loop:
+	mvc	0(256,%r1),0(%r3)
+	la	%r1,256(%r1)
+	la	%r3,256(%r3)
+	brct	%r0,.Lmemcpy_loop
+	j	.Lmemcpy_rest
+.Lmemcpy_mvc:
+	mvc	0(1,%r1),0(%r3)
diff --git a/arch/s390/lib/mem64.S b/arch/s390/lib/mem64.S
new file mode 100644
index 0000000..c6d553e
--- /dev/null
+++ b/arch/s390/lib/mem64.S
@@ -0,0 +1,88 @@
+/*
+ * String handling functions.
+ *
+ * Copyright IBM Corp. 2012
+ */
+
+#include <linux/linkage.h>
+
+/*
+ * memset implementation
+ *
+ * This code corresponds to the C construct below. We do distinguish
+ * between clearing (c == 0) and setting a memory array (c != 0) simply
+ * because nearly all memset invocations in the kernel clear memory and
+ * the xc instruction is preferred in such cases.
+ *
+ * void *memset(void *s, int c, size_t n)
+ * {
+ *	if (likely(c == 0))
+ *		return __builtin_memset(s, 0, n);
+ *	return __builtin_memset(s, c, n);
+ * }
+ */
+ENTRY(memset)
+	ltgr	%r4,%r4
+	bzr	%r14
+	ltgr	%r3,%r3
+	jnz	.Lmemset_fill
+	aghi	%r4,-1
+	srlg	%r3,%r4,8
+	ltgr	%r3,%r3
+	lgr	%r1,%r2
+	jz	.Lmemset_clear_rest
+.Lmemset_clear_loop:
+	xc	0(256,%r1),0(%r1)
+	la	%r1,256(%r1)
+	brctg	%r3,.Lmemset_clear_loop
+.Lmemset_clear_rest:
+	larl	%r3,.Lmemset_xc
+	ex	%r4,0(%r3)
+	br	%r14
+.Lmemset_fill:
+	stc	%r3,0(%r2)
+	cghi	%r4,1
+	lgr	%r1,%r2
+	ber	%r14
+	aghi	%r4,-2
+	srlg	%r3,%r4,8
+	ltgr	%r3,%r3
+	jz	.Lmemset_fill_rest
+.Lmemset_fill_loop:
+	mvc	1(256,%r1),0(%r1)
+	la	%r1,256(%r1)
+	brctg	%r3,.Lmemset_fill_loop
+.Lmemset_fill_rest:
+	larl	%r3,.Lmemset_mvc
+	ex	%r4,0(%r3)
+	br	%r14
+.Lmemset_xc:
+	xc	0(1,%r1),0(%r1)
+.Lmemset_mvc:
+	mvc	1(1,%r1),0(%r1)
+
+/*
+ * memcpy implementation
+ *
+ * void *memcpy(void *dest, const void *src, size_t n)
+ */
+ENTRY(memcpy)
+	ltgr	%r4,%r4
+	bzr	%r14
+	aghi	%r4,-1
+	srlg	%r5,%r4,8
+	ltgr	%r5,%r5
+	lgr	%r1,%r2
+	jnz	.Lmemcpy_loop
+.Lmemcpy_rest:
+	larl	%r5,.Lmemcpy_mvc
+	ex	%r4,0(%r5)
+	br	%r14
+.Lmemcpy_loop:
+	mvc	0(256,%r1),0(%r3)
+	la	%r1,256(%r1)
+	la	%r3,256(%r3)
+	brctg	%r5,.Lmemcpy_loop
+	j	.Lmemcpy_rest
+.Lmemcpy_mvc:
+	mvc	0(1,%r1),0(%r3)
diff --git a/arch/s390/lib/string.c b/arch/s390/lib/string.c
index 846ec64..b647d5f 100644
--- a/arch/s390/lib/string.c
+++ b/arch/s390/lib/string.c
@@ -43,11 +43,7 @@ static inline char *__strnend(const char *s, size_t n)
  */
 size_t strlen(const char *s)
 {
-#if __GNUC__ < 4
 	return __strend(s) - s;
-#else
-	return __builtin_strlen(s);
-#endif
 }
 EXPORT_SYMBOL(strlen);
 
@@ -73,7 +69,6 @@ EXPORT_SYMBOL(strnlen);
  */
 char *strcpy(char *dest, const char *src)
 {
-#if __GNUC__ < 4
 	register int r0 asm("0") = 0;
 	char *ret = dest;
 
@@ -82,9 +77,6 @@ char *strcpy(char *dest, const char *src)
 		      : "+&a" (dest), "+&a" (src) : "d" (r0)
 		      : "cc", "memory" );
 	return ret;
-#else
-	return __builtin_strcpy(dest, src);
-#endif
 }
 EXPORT_SYMBOL(strcpy);
 
@@ -106,7 +98,7 @@ size_t strlcpy(char *dest, const char *src, size_t size)
 	if (size) {
 		size_t len = (ret >= size) ? size-1 : ret;
 		dest[len] = '\0';
-		__builtin_memcpy(dest, src, len);
+		memcpy(dest, src, len);
 	}
 	return ret;
 }
@@ -124,8 +116,8 @@ EXPORT_SYMBOL(strlcpy);
 char *strncpy(char *dest, const char *src, size_t n)
 {
 	size_t len = __strnend(src, n) - src;
-	__builtin_memset(dest + len, 0, n - len);
-	__builtin_memcpy(dest, src, len);
+	memset(dest + len, 0, n - len);
+	memcpy(dest, src, len);
 	return dest;
 }
 EXPORT_SYMBOL(strncpy);
@@ -171,7 +163,7 @@ size_t strlcat(char *dest, const char *src, size_t n)
 		if (len >= n)
 			len = n - 1;
 		dest[len] = '\0';
-		__builtin_memcpy(dest, src, len);
+		memcpy(dest, src, len);
 	}
 	return res;
 }
@@ -194,7 +186,7 @@ char *strncat(char *dest, const char *src, size_t n)
 	char *p = __strend(dest);
 
 	p[len] = '\0';
-	__builtin_memcpy(p, src, len);
+	memcpy(p, src, len);
 	return dest;
 }
 EXPORT_SYMBOL(strncat);
@@ -348,41 +340,3 @@ void *memscan(void *s, int c, size_t n)
 	return (void *) ret;
 }
 EXPORT_SYMBOL(memscan);
-
-/**
- * memcpy - Copy one area of memory to another
- * @dest: Where to copy to
- * @src: Where to copy from
- * @n: The size of the area.
- *
- * returns a pointer to @dest
- */
-void *memcpy(void *dest, const void *src, size_t n)
-{
-	return __builtin_memcpy(dest, src, n);
-}
-EXPORT_SYMBOL(memcpy);
-
-/**
- * memset - Fill a region of memory with the given value
- * @s: Pointer to the start of the area.
- * @c: The byte to fill the area with
- * @n: The size of the area.
- *
- * returns a pointer to @s
- */
-void *memset(void *s, int c, size_t n)
-{
-	char *xs;
-
-	if (c == 0)
-		return __builtin_memset(s, 0, n);
-
-	xs = (char *) s;
-	if (n > 0)
-		do {
-			*xs++ = c;
-		} while (--n > 0);
-	return s;
-}
-EXPORT_SYMBOL(memset);