PPC assembler pack: adhere closer to ABI specs, add PowerOpen traceback data.

2 files changed, 101 insertions, 96 deletions

diff --git a/crypto/sha/asm/sha1-ppc.pl b/crypto/sha/asm/sha1-ppc.pl
index dcd0fcdfcf..2140dd2f8d 100755
--- a/crypto/sha/asm/sha1-ppc.pl
+++ b/crypto/sha/asm/sha1-ppc.pl
@@ -24,12 +24,14 @@ $flavour = shift;
 
 if ($flavour =~ /64/) {
 	$SIZE_T	=8;
+	$LRSAVE	=2*$SIZE_T;
 	$UCMP	="cmpld";
 	$STU	="stdu";
 	$POP	="ld";
 	$PUSH	="std";
 } elsif ($flavour =~ /32/) {
 	$SIZE_T	=4;
+	$LRSAVE	=$SIZE_T;
 	$UCMP	="cmplw";
 	$STU	="stwu";
 	$POP	="lwz";
@@ -43,7 +45,8 @@ die "can't locate ppc-xlate.pl";
 
 open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
 
-$FRAME=24*$SIZE_T;
+$FRAME=24*$SIZE_T+64;
+$LOCALS=6*$SIZE_T;
 
 $K  ="r0";
 $sp ="r1";
@@ -162,9 +165,8 @@ $code=<<___;
 .globl	.sha1_block_data_order
 .align	4
 .sha1_block_data_order:
+	$STU	$sp,-$FRAME($sp)
 	mflr	r0
-	$STU	$sp,`-($FRAME+64)`($sp)
-	$PUSH	r0,`$FRAME-$SIZE_T*18`($sp)
 	$PUSH	r15,`$FRAME-$SIZE_T*17`($sp)
 	$PUSH	r16,`$FRAME-$SIZE_T*16`($sp)
 	$PUSH	r17,`$FRAME-$SIZE_T*15`($sp)
@@ -182,6 +184,7 @@ $code=<<___;
 	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
+	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
 	lwz	$A,0($ctx)
 	lwz	$B,4($ctx)
 	lwz	$C,8($ctx)
@@ -192,37 +195,14 @@ $code=<<___;
 Laligned:
 	mtctr	$num
 	bl	Lsha1_block_private
-Ldone:
-	$POP	r0,`$FRAME-$SIZE_T*18`($sp)
-	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
-	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
-	$POP	r17,`$FRAME-$SIZE_T*15`($sp)
-	$POP	r18,`$FRAME-$SIZE_T*14`($sp)
-	$POP	r19,`$FRAME-$SIZE_T*13`($sp)
-	$POP	r20,`$FRAME-$SIZE_T*12`($sp)
-	$POP	r21,`$FRAME-$SIZE_T*11`($sp)
-	$POP	r22,`$FRAME-$SIZE_T*10`($sp)
-	$POP	r23,`$FRAME-$SIZE_T*9`($sp)
-	$POP	r24,`$FRAME-$SIZE_T*8`($sp)
-	$POP	r25,`$FRAME-$SIZE_T*7`($sp)
-	$POP	r26,`$FRAME-$SIZE_T*6`($sp)
-	$POP	r27,`$FRAME-$SIZE_T*5`($sp)
-	$POP	r28,`$FRAME-$SIZE_T*4`($sp)
-	$POP	r29,`$FRAME-$SIZE_T*3`($sp)
-	$POP	r30,`$FRAME-$SIZE_T*2`($sp)
-	$POP	r31,`$FRAME-$SIZE_T*1`($sp)
-	mtlr	r0
-	addi	$sp,$sp,`$FRAME+64`
-	blr
-___
+	b	Ldone
 
-# PowerPC specification allows an implementation to be ill-behaved
-# upon unaligned access which crosses page boundary. "Better safe
-# than sorry" principle makes me treat it specially. But I don't
-# look for particular offending word, but rather for 64-byte input
-# block which crosses the boundary. Once found that block is aligned
-# and hashed separately...
-$code.=<<___;
+; PowerPC specification allows an implementation to be ill-behaved
+; upon unaligned access which crosses page boundary. "Better safe
+; than sorry" principle makes me treat it specially. But I don't
+; look for particular offending word, but rather for 64-byte input
+; block which crosses the boundary. Once found that block is aligned
+; and hashed separately...
 .align	4
 Lunaligned:
 	subfic	$t1,$inp,4096
@@ -237,7 +217,7 @@ Lunaligned:
 Lcross_page:
 	li	$t1,16
 	mtctr	$t1
-	addi	r20,$sp,$FRAME	; spot below the frame
+	addi	r20,$sp,$LOCALS	; spot within the frame
 Lmemcpy:
 	lbz	r16,0($inp)
 	lbz	r17,1($inp)
@@ -251,15 +231,40 @@ Lmemcpy:
 	addi	r20,r20,4
 	bdnz	Lmemcpy
 
-	$PUSH	$inp,`$FRAME-$SIZE_T*19`($sp)
+	$PUSH	$inp,`$FRAME-$SIZE_T*18`($sp)
 	li	$t1,1
-	addi	$inp,$sp,$FRAME
+	addi	$inp,$sp,$LOCALS
 	mtctr	$t1
 	bl	Lsha1_block_private
-	$POP	$inp,`$FRAME-$SIZE_T*19`($sp)
+	$POP	$inp,`$FRAME-$SIZE_T*18`($sp)
 	addic.	$num,$num,-1
 	bne-	Lunaligned
-	b	Ldone
+
+Ldone:
+	$POP	r0,`$FRAME+$LRSAVE`($sp)
+	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
+	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
+	$POP	r17,`$FRAME-$SIZE_T*15`($sp)
+	$POP	r18,`$FRAME-$SIZE_T*14`($sp)
+	$POP	r19,`$FRAME-$SIZE_T*13`($sp)
+	$POP	r20,`$FRAME-$SIZE_T*12`($sp)
+	$POP	r21,`$FRAME-$SIZE_T*11`($sp)
+	$POP	r22,`$FRAME-$SIZE_T*10`($sp)
+	$POP	r23,`$FRAME-$SIZE_T*9`($sp)
+	$POP	r24,`$FRAME-$SIZE_T*8`($sp)
+	$POP	r25,`$FRAME-$SIZE_T*7`($sp)
+	$POP	r26,`$FRAME-$SIZE_T*6`($sp)
+	$POP	r27,`$FRAME-$SIZE_T*5`($sp)
+	$POP	r28,`$FRAME-$SIZE_T*4`($sp)
+	$POP	r29,`$FRAME-$SIZE_T*3`($sp)
+	$POP	r30,`$FRAME-$SIZE_T*2`($sp)
+	$POP	r31,`$FRAME-$SIZE_T*1`($sp)
+	mtlr	r0
+	addi	$sp,$sp,$FRAME
+	blr
+	.long	0
+	.byte	0,12,4,1,0x80,18,3,0
+	.long	0
 ___
 
 # This is private block function, which uses tailored calling
@@ -309,6 +314,8 @@ $code.=<<___;
 	addi	$inp,$inp,`16*4`
 	bdnz-	Lsha1_block_private
 	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
 ___
 $code.=<<___;
 .asciz	"SHA1 block transform for PPC, CRYPTOGAMS by <appro\@fy.chalmers.se>"
diff --git a/crypto/sha/asm/sha512-ppc.pl b/crypto/sha/asm/sha512-ppc.pl
index 768a6a6fad..6b44a68e59 100755
--- a/crypto/sha/asm/sha512-ppc.pl
+++ b/crypto/sha/asm/sha512-ppc.pl
@@ -40,6 +40,7 @@ $output =shift;
 
 if ($flavour =~ /64/) {
 	$SIZE_T=8;
+	$LRSAVE=2*$SIZE_T;
 	$STU="stdu";
 	$UCMP="cmpld";
 	$SHL="sldi";
@@ -47,6 +48,7 @@ if ($flavour =~ /64/) {
 	$PUSH="std";
 } elsif ($flavour =~ /32/) {
 	$SIZE_T=4;
+	$LRSAVE=$SIZE_T;
 	$STU="stwu";
 	$UCMP="cmplw";
 	$SHL="slwi";
@@ -87,7 +89,8 @@ if ($output =~ /512/) {
 	$SHR="srwi";
 }
 
-$FRAME=32*$SIZE_T;
+$FRAME=32*$SIZE_T+16*$SZ;
+$LOCALS=6*$SIZE_T;
 
 $sp ="r1";
 $toc="r2";
@@ -179,13 +182,12 @@ $code=<<___;
 .globl	$func
 .align	6
 $func:
+	$STU	$sp,-$FRAME($sp)
 	mflr	r0
-	$STU	$sp,`-($FRAME+16*$SZ)`($sp)
 	$SHL	$num,$num,`log(16*$SZ)/log(2)`
 
 	$PUSH	$ctx,`$FRAME-$SIZE_T*22`($sp)
 
-	$PUSH	r0,`$FRAME-$SIZE_T*21`($sp)
 	$PUSH	$toc,`$FRAME-$SIZE_T*20`($sp)
 	$PUSH	r13,`$FRAME-$SIZE_T*19`($sp)
 	$PUSH	r14,`$FRAME-$SIZE_T*18`($sp)
@@ -206,6 +208,7 @@ $func:
 	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
+	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
 
 	$LD	$A,`0*$SZ`($ctx)
 	mr	$inp,r4				; incarnate $inp
@@ -217,7 +220,7 @@ $func:
 	$LD	$G,`6*$SZ`($ctx)
 	$LD	$H,`7*$SZ`($ctx)
 
-	b	LPICmeup
+	bl	LPICmeup
 LPICedup:
 	andi.	r0,$inp,3
 	bne	Lunaligned
@@ -226,40 +229,14 @@ Laligned:
 	$PUSH	$num,`$FRAME-$SIZE_T*24`($sp)	; end pointer
 	$PUSH	$inp,`$FRAME-$SIZE_T*23`($sp)	; inp pointer
 	bl	Lsha2_block_private
-Ldone:
-	$POP	r0,`$FRAME-$SIZE_T*21`($sp)
-	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
-	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
-	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
-	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
-	$POP	r17,`$FRAME-$SIZE_T*15`($sp)
-	$POP	r18,`$FRAME-$SIZE_T*14`($sp)
-	$POP	r19,`$FRAME-$SIZE_T*13`($sp)
-	$POP	r20,`$FRAME-$SIZE_T*12`($sp)
-	$POP	r21,`$FRAME-$SIZE_T*11`($sp)
-	$POP	r22,`$FRAME-$SIZE_T*10`($sp)
-	$POP	r23,`$FRAME-$SIZE_T*9`($sp)
-	$POP	r24,`$FRAME-$SIZE_T*8`($sp)
-	$POP	r25,`$FRAME-$SIZE_T*7`($sp)
-	$POP	r26,`$FRAME-$SIZE_T*6`($sp)
-	$POP	r27,`$FRAME-$SIZE_T*5`($sp)
-	$POP	r28,`$FRAME-$SIZE_T*4`($sp)
-	$POP	r29,`$FRAME-$SIZE_T*3`($sp)
-	$POP	r30,`$FRAME-$SIZE_T*2`($sp)
-	$POP	r31,`$FRAME-$SIZE_T*1`($sp)
-	mtlr	r0
-	addi	$sp,$sp,`$FRAME+16*$SZ`
-	blr
-___
+	b	Ldone
 
-# PowerPC specification allows an implementation to be ill-behaved
-# upon unaligned access which crosses page boundary. "Better safe
-# than sorry" principle makes me treat it specially. But I don't
-# look for particular offending word, but rather for the input
-# block which crosses the boundary. Once found that block is aligned
-# and hashed separately...
-$code.=<<___;
+; PowerPC specification allows an implementation to be ill-behaved
+; upon unaligned access which crosses page boundary. "Better safe
+; than sorry" principle makes me treat it specially. But I don't
+; look for particular offending word, but rather for the input
+; block which crosses the boundary. Once found that block is aligned
+; and hashed separately...
 .align	4
 Lunaligned:
 	subfic	$t1,$inp,4096
@@ -278,7 +255,7 @@ Lunaligned:
 Lcross_page:
 	li	$t1,`16*$SZ/4`
 	mtctr	$t1
-	addi	r20,$sp,$FRAME			; aligned spot below the frame
+	addi	r20,$sp,$LOCALS			; aligned spot below the frame
 Lmemcpy:
 	lbz	r16,0($inp)
 	lbz	r17,1($inp)
@@ -293,8 +270,8 @@ Lmemcpy:
 	bdnz	Lmemcpy
 
 	$PUSH	$inp,`$FRAME-$SIZE_T*26`($sp)	; save real inp
-	addi	$t1,$sp,`$FRAME+16*$SZ`		; fictitious end pointer
-	addi	$inp,$sp,$FRAME			; fictitious inp pointer
+	addi	$t1,$sp,`$LOCALS+16*$SZ`	; fictitious end pointer
+	addi	$inp,$sp,$LOCALS		; fictitious inp pointer
 	$PUSH	$num,`$FRAME-$SIZE_T*25`($sp)	; save real num
 	$PUSH	$t1,`$FRAME-$SIZE_T*24`($sp)	; end pointer
 	$PUSH	$inp,`$FRAME-$SIZE_T*23`($sp)	; inp pointer
@@ -303,10 +280,36 @@ Lmemcpy:
 	$POP	$num,`$FRAME-$SIZE_T*25`($sp)	; restore real num
 	addic.	$num,$num,`-16*$SZ`		; num--
 	bne-	Lunaligned
-	b	Ldone
-___
 
-$code.=<<___;
+Ldone:
+	$POP	r0,`$FRAME+$LRSAVE`($sp)
+	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
+	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
+	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
+	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
+	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
+	$POP	r17,`$FRAME-$SIZE_T*15`($sp)
+	$POP	r18,`$FRAME-$SIZE_T*14`($sp)
+	$POP	r19,`$FRAME-$SIZE_T*13`($sp)
+	$POP	r20,`$FRAME-$SIZE_T*12`($sp)
+	$POP	r21,`$FRAME-$SIZE_T*11`($sp)
+	$POP	r22,`$FRAME-$SIZE_T*10`($sp)
+	$POP	r23,`$FRAME-$SIZE_T*9`($sp)
+	$POP	r24,`$FRAME-$SIZE_T*8`($sp)
+	$POP	r25,`$FRAME-$SIZE_T*7`($sp)
+	$POP	r26,`$FRAME-$SIZE_T*6`($sp)
+	$POP	r27,`$FRAME-$SIZE_T*5`($sp)
+	$POP	r28,`$FRAME-$SIZE_T*4`($sp)
+	$POP	r29,`$FRAME-$SIZE_T*3`($sp)
+	$POP	r30,`$FRAME-$SIZE_T*2`($sp)
+	$POP	r31,`$FRAME-$SIZE_T*1`($sp)
+	mtlr	r0
+	addi	$sp,$sp,$FRAME
+	blr
+	.long	0
+	.byte	0,12,4,1,0x80,18,3,0
+	.long	0
+
 .align	4
 Lsha2_block_private:
 ___
@@ -372,6 +375,8 @@ $code.=<<___;
 	$ST	$H,`7*$SZ`($ctx)
 	bne	Lsha2_block_private
 	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
 ___
 
 # Ugly hack here, because PPC assembler syntax seem to vary too
@@ -379,22 +384,15 @@ ___
 $code.=<<___;
 .align	6
 LPICmeup:
-	bl	LPIC
-	addi	$Tbl,$Tbl,`64-4`	; "distance" between . and last nop
-	b	LPICedup
-	nop
-	nop
-	nop
-	nop
-	nop
-LPIC:	mflr	$Tbl
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	$Tbl	; vvvvvv "distance" between . and 1st data entry
+	addi	$Tbl,$Tbl,`64-8`
+	mtlr	r0
 	blr
-	nop
-	nop
-	nop
-	nop
-	nop
-	nop
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+	.space	`64-9*4`
 ___
 $code.=<<___ if ($SZ==8);
 	.long	0x428a2f98,0xd728ae22,0x71374491,0x23ef65cd