patch-2.1.29 linux/arch/sparc64/lib/memset.S

• Lines: 268
• Date: Wed Mar 5 17:04:31 1997
• Orig file: v2.1.28/linux/arch/sparc64/lib/memset.S
• Orig date: Mon Dec 30 01:59:59 1996

diff -u --recursive --new-file v2.1.28/linux/arch/sparc64/lib/memset.S linux/arch/sparc64/lib/memset.S
@@ -1,12 +1,36 @@
-/* $Id: memset.S,v 1.1 1996/12/22 07:42:16 davem Exp $
- * arch/sparc64/lib/memset.S: UltraSparc optimized memset and bzero code
- *
+/* linux/arch/sparc64/lib/memset.S: Sparc optimized memset, bzero and clear_user code
+ * Copyright (C) 1991,1996 Free Software Foundation
  * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
+ * Copyright (C) 1996,1997 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
+ *
+ * Returns 0, if ok, and number of bytes not yet set if exception
+ * occurs and we were called as clear_user.
  */
 
-#include <asm/asi.h>
+#include <asm/ptrace.h>
 
-	/* Store 64 bytes at (BASE + OFFSET) using value SOURCE. */
+#define EX(x,y,a,b,z) 				\
+98: 	x,y;					\
+	.section .fixup,z##alloc,z##execinstr;	\
+	.align	4;				\
+99:	ba,pt	%xcc, 30f;			\
+	 a, b, %o0;				\
+	.section __ex_table,z##alloc;		\
+	.align	4;				\
+	.word	98b, 99b;			\
+	.text;					\
+	.align	4
+
+#define EXT(start,end,handler,z) 		\
+	.section __ex_table,z##alloc;		\
+	.align	4;				\
+	.word	start, 0, end, handler;		\
+	.text;					\
+	.align	4
+
+/* Please don't change these macros, unless you change the logic
+ * in the .fixup section below as well.
+ * Store 64 bytes at (BASE + OFFSET) using value SOURCE. */
 #define ZERO_BIG_BLOCK(base, offset, source)    \
 	stx	source, [base + offset + 0x00]; \
 	stx	source, [base + offset + 0x08]; \
@@ -30,104 +54,139 @@
 	.text
 	.align 4
 
-	.globl	C_LABEL(__bzero), C_LABEL(__memset), C_LABEL(memset)
-C_LABEL(__memset):
-C_LABEL(memset):
-	and		%o1, 0xff, %g3
-	sll		%g3, 8, %g2
-	or		%g3, %g2, %g3
-	sll		%g3, 16, %g2
-	or		%g3, %g2, %g3
-	sllx		%g3, 32, %g2
-	or		%g3, %g2, %g3
-	b		1f
-	 mov		%o2, %o1
-
-3:
-	cmp		%o2, 3
-	be		2f
-	 stb		%g3, [%o0]
-
-	cmp		%o2, 2
-	be		2f
-	 stb		%g3, [%o0 + 0x01]
-
-	stb		%g3, [%o0 + 0x02]
-2:
-	sub		%o2, 4, %o2
-	add		%o1, %o2, %o1
-	b		4f
-	 sub		%o0, %o2, %o0
-
-C_LABEL(__bzero):
-	mov		%g0, %g3
+	.globl	__bzero, __memset, 
+	.globl	memset, __memset_start, __memset_end
+__memset_start:
+__memset:
+memset:
+	and	%o1, 0xff, %g3
+	sll	%g3, 8, %g2
+	or	%g3, %g2, %g3
+	sll	%g3, 16, %g2
+	or	%g3, %g2, %g3
+	mov	%o2, %o1
+	sllx	%g3, 32, %g2
+	ba,pt	%xcc, 1f
+	 or	%g3, %g2, %g3
+__bzero:
+	mov	%g0, %g3
 1:
-	cmp		%o1, 7
-	bleu,pnt	%icc, 7f
-	 mov		%o0, %g1
+	cmp	%o1, 7
+	bleu,pn	%xcc, 7f
+	 andcc	%o0, 3, %o2
+
+	be,a,pt	%icc, 4f
+	 andcc	%o0, 4, %g0
+
+	cmp	%o2, 3
+	be,pn	%icc, 2f
+	 EX(stb	%g3, [%o0], sub %o1, 0,#)
+
+	cmp	%o2, 2
+	be,pt	%icc, 2f
+	 EX(stb	%g3, [%o0 + 0x01], sub %o1, 1,#)
 
-	andcc		%o0, 3, %o2
-	bne,pnt		%icc, 3b
+	EX(stb	%g3, [%o0 + 0x02], sub %o1, 2,#)
+2:
+	sub	%o2, 4, %o2
+	sub	%o0, %o2, %o0
+	add	%o1, %o2, %o1
+	andcc	%o0, 4, %g0
 4:
-	 andcc		%o0, 4, %g0
+	be,a,pt	%icc, 2f
+	 andncc	%o1, 0x7f, %o3
 
-	be,a,pt		%icc, 2f
-	 andcc		%o1, 0xffffff80, %o3	! everything 8 aligned, o1 is len to run
-
-	stw		%g3, [%o0]
-	sub		%o1, 4, %o1
-	add		%o0, 4, %o0
-	andcc		%o1, 0xffffff80, %o3	! everything 8 aligned, o1 is len to run
+	EX(st	%g3, [%o0], sub %o1, 0,#)
+	sub	%o1, 4, %o1
+	add	%o0, 4, %o0
+	andncc	%o1, 0x7f, %o3		! Now everything is 8 aligned and o1 is len to run
 2:
-	be		9f
-	 andcc		%o1, 0x78, %o2
-4:
-	ZERO_BIG_BLOCK(%o0, 0x00, %g2)
-	subcc		%o3, 128, %o3
-	ZERO_BIG_BLOCK(%o0, 0x40, %g2)
-	bne,pt		%icc, 4b
-	 add		%o0, 128, %o0
+	be,pn	%xcc, 9f
+	 andcc	%o1, 0x78, %o2
+10:
+	ZERO_BIG_BLOCK(%o0, 0x00, %g3)
+	subcc	%o3, 128, %o3
+	ZERO_BIG_BLOCK(%o0, 0x40, %g3)
+11:
+	EXT(10b, 11b, 20f,#)
+	bne,pt	%xcc, 10b
+	 add	%o0, 128, %o0
 
-	orcc		%o2, %g0, %g0
+	tst	%o2
 9:
-	be,pnt		%icc, 6f
-	 andcc		%o1, 7, %o1
+	be,pn	%xcc, 13f
+	 andcc	%o1, 7, %o1
+14:
+	rd	%pc, %o4
+	srl	%o2, 1, %o3
+	sub	%o4, %o3, %o4
+	jmpl	%o4 + (13f - 14b), %g0
+	 add	%o0, %o2, %o0
+12:
+	ZERO_LAST_BLOCKS(%o0, 0x48, %g3)
+	ZERO_LAST_BLOCKS(%o0, 0x08, %g3)
+13:
+	be,pn	%icc, 8f
+	 andcc	%o1, 4, %g0
 
-	srl		%o2, 1, %o3
-	set		bzero_table + 64, %o4
-	sub		%o4, %o3, %o4
-	jmp		%o4
-	 add		%o0, %o2, %o0
-
-bzero_table:
-	ZERO_LAST_BLOCKS(%o0, 0x48, %g2)
-	ZERO_LAST_BLOCKS(%o0, 0x08, %g2)
-
-6:
-	be,pt		%icc, 8f
-	 andcc		%o1, 4, %g0
+	be,pn	%icc, 1f
+	 andcc	%o1, 2, %g0
 
-	be,pnt		%icc, 1f
-	 andcc		%o1, 2, %g0
-
-	stw		%g3, [%o0]
-	add		%o0, 4, %o0
+	EX(st	%g3, [%o0], and %o1, 7,#)
+	add	%o0, 4, %o0
 1:
-	be,pt		%icc, 1f
-	 andcc		%o1, 1, %g0
+	be,pn	%icc, 1f
+	 andcc	%o1, 1, %g0
 
-	sth		%g3, [%o0]
-	add		%o0, 2, %o0
+	EX(sth	%g3, [%o0], and %o1, 3,#)
+	add	%o0, 2, %o0
 1:
-	bne,a,pnt	%icc, 8f
-	 stb		%g3, [%o0]
+	bne,a,pn %icc, 8f
+	 EX(stb	%g3, [%o0], and %o1, 1,#)
 8:
 	retl
-	 mov		%g1, %o0
-
-/* Don't care about alignment here. It is highly 
- * unprobable and at most two traps may happen
- */
+	 clr	%o0
 7:
-	ba,pt		%xcc, 6b
-	 orcc		%o1, 0, %g0
+	be,pn	%icc, 13b
+	 orcc	%o1, 0, %g0
+
+	be,pn	%icc, 0f
+8:
+	 add	%o0, 1, %o0
+	subcc	%o1, 1, %o1
+	bne,a,pt %icc, 8b
+	 EX(stb	%g3, [%o0 - 1], add %o1, 1,#)
+0:
+	retl
+	 clr	%o0
+__memset_end:
+
+	.section .fixup,#alloc,#execinstr
+	.align	4
+20:
+	cmp	%g2, 8
+	bleu,pn	%xcc, 1f
+	 and	%o1, 0x7f, %o1
+	sub	%g2, 9, %g2
+	add	%o3, 64, %o3
+1:
+	sll	%g2, 3, %g2
+	add	%o3, %o1, %o0
+	ba,pt	%xcc, 30f
+	 sub	%o0, %g2, %o0
+21:
+	mov	8, %o0
+	and	%o1, 7, %o1
+	sub	%o0, %g2, %o0
+	sll	%o0, 3, %o0
+	ba,pt	%xcc, 30f
+	 add	%o0, %o1, %o0
+30:
+/* %o4 is faulting address, %o5 is %pc where fault occured */
+	save	%sp, -160, %sp
+	mov	%i5, %o0
+	mov	%i7, %o1
+	call	lookup_fault
+	 mov	%i4, %o2
+	ret
+	 restore