Fix CVE-2014-3570: properly calculate the square of a BIGNUM value.

See https://www.openssl.org/news/secadv_20150108.txt for a more detailed
discussion.

Original OpenSSL patch here:
https://github.com/openssl/openssl/commit/a7a44ba55cb4f884c6bc9ceac90072dea38e66d0

The regression test is modified a little for KNF.
ok miod@

3 files changed, 325 insertions, 678 deletions

diff --git a/lib/libssl/src/crypto/bn/asm/mips.pl b/lib/libssl/src/crypto/bn/asm/mips.pl
index c162a3ec230..215c9a74832 100644
--- a/lib/libssl/src/crypto/bn/asm/mips.pl
+++ b/lib/libssl/src/crypto/bn/asm/mips.pl
@@ -140,10 +140,10 @@ $code.=<<___;
 	.set	reorder
 	li	$minus4,-4
 	and	$ta0,$a2,$minus4
-	$LD	$t0,0($a1)
 	beqz	$ta0,.L_bn_mul_add_words_tail
 
 .L_bn_mul_add_words_loop:
+	$LD	$t0,0($a1)
 	$MULTU	$t0,$a3
 	$LD	$t1,0($a0)
 	$LD	$t2,$BNSZ($a1)
@@ -200,10 +200,9 @@ $code.=<<___;
 	$ADDU	$v0,$ta2
 	sltu	$at,$ta3,$at
 	$ST	$ta3,-$BNSZ($a0)
-	$ADDU	$v0,$at
 	.set	noreorder
-	bgtzl	$ta0,.L_bn_mul_add_words_loop
-	$LD	$t0,0($a1)
+	bgtz	$ta0,.L_bn_mul_add_words_loop
+	$ADDU	$v0,$at
 
 	beqz	$a2,.L_bn_mul_add_words_return
 	nop
@@ -300,10 +299,10 @@ $code.=<<___;
 	.set	reorder
 	li	$minus4,-4
 	and	$ta0,$a2,$minus4
-	$LD	$t0,0($a1)
 	beqz	$ta0,.L_bn_mul_words_tail
 
 .L_bn_mul_words_loop:
+	$LD	$t0,0($a1)
 	$MULTU	$t0,$a3
 	$LD	$t2,$BNSZ($a1)
 	$LD	$ta0,2*$BNSZ($a1)
@@ -341,10 +340,9 @@ $code.=<<___;
 	$ADDU	$v0,$at
 	sltu	$ta3,$v0,$at
 	$ST	$v0,-$BNSZ($a0)
-	$ADDU	$v0,$ta3,$ta2
 	.set	noreorder
-	bgtzl	$ta0,.L_bn_mul_words_loop
-	$LD	$t0,0($a1)
+	bgtz	$ta0,.L_bn_mul_words_loop
+	$ADDU	$v0,$ta3,$ta2
 
 	beqz	$a2,.L_bn_mul_words_return
 	nop
@@ -429,10 +427,10 @@ $code.=<<___;
 	.set	reorder
 	li	$minus4,-4
 	and	$ta0,$a2,$minus4
-	$LD	$t0,0($a1)
 	beqz	$ta0,.L_bn_sqr_words_tail
 
 .L_bn_sqr_words_loop:
+	$LD	$t0,0($a1)
 	$MULTU	$t0,$t0
 	$LD	$t2,$BNSZ($a1)
 	$LD	$ta0,2*$BNSZ($a1)
@@ -463,11 +461,10 @@ $code.=<<___;
 	mflo	$ta3
 	mfhi	$ta2
 	$ST	$ta3,-2*$BNSZ($a0)
-	$ST	$ta2,-$BNSZ($a0)
 
 	.set	noreorder
-	bgtzl	$ta0,.L_bn_sqr_words_loop
-	$LD	$t0,0($a1)
+	bgtz	$ta0,.L_bn_sqr_words_loop
+	$ST	$ta2,-$BNSZ($a0)
 
 	beqz	$a2,.L_bn_sqr_words_return
 	nop
@@ -547,10 +544,10 @@ $code.=<<___;
 	.set	reorder
 	li	$minus4,-4
 	and	$at,$a3,$minus4
-	$LD	$t0,0($a1)
 	beqz	$at,.L_bn_add_words_tail
 
 .L_bn_add_words_loop:
+	$LD	$t0,0($a1)
 	$LD	$ta0,0($a2)
 	subu	$a3,4
 	$LD	$t1,$BNSZ($a1)
@@ -589,11 +586,10 @@ $code.=<<___;
 	$ADDU	$t3,$ta3,$v0
 	sltu	$v0,$t3,$ta3
 	$ST	$t3,-$BNSZ($a0)
-	$ADDU	$v0,$t9
 	
 	.set	noreorder
-	bgtzl	$at,.L_bn_add_words_loop
-	$LD	$t0,0($a1)
+	bgtz	$at,.L_bn_add_words_loop
+	$ADDU	$v0,$t9
 
 	beqz	$a3,.L_bn_add_words_return
 	nop
@@ -679,10 +675,10 @@ $code.=<<___;
 	.set	reorder
 	li	$minus4,-4
 	and	$at,$a3,$minus4
-	$LD	$t0,0($a1)
 	beqz	$at,.L_bn_sub_words_tail
 
 .L_bn_sub_words_loop:
+	$LD	$t0,0($a1)
 	$LD	$ta0,0($a2)
 	subu	$a3,4
 	$LD	$t1,$BNSZ($a1)
@@ -722,11 +718,10 @@ $code.=<<___;
 	$SUBU	$t3,$ta3,$v0
 	sgtu	$v0,$t3,$ta3
 	$ST	$t3,-$BNSZ($a0)
-	$ADDU	$v0,$t9
 
 	.set	noreorder
-	bgtzl	$at,.L_bn_sub_words_loop
-	$LD	$t0,0($a1)
+	bgtz	$at,.L_bn_sub_words_loop
+	$ADDU	$v0,$t9
 
 	beqz	$a3,.L_bn_sub_words_return
 	nop
@@ -819,7 +814,7 @@ ___
 $code.=<<___;
 	.set	reorder
 	move	$ta3,$ra
-	bal	bn_div_words
+	bal	bn_div_words_internal
 	move	$ra,$ta3
 	$MULTU	$ta2,$v0
 	$LD	$t2,-2*$BNSZ($a3)
@@ -840,8 +835,9 @@ $code.=<<___;
 	sltu	$ta0,$a1,$a2
 	or	$t8,$ta0
 	.set	noreorder
-	beqzl	$at,.L_bn_div_3_words_inner_loop
+	beqz	$at,.L_bn_div_3_words_inner_loop
 	$SUBU	$v0,1
+	$ADDU	$v0,1
 	.set	reorder
 .L_bn_div_3_words_inner_loop_done:
 	.set	noreorder
@@ -902,7 +898,8 @@ $code.=<<___;
 	and	$t2,$a0
 	$SRL	$at,$a1,$t1
 	.set	noreorder
-	bnezl	$t2,.+8
+	beqz	$t2,.+12
+	nop
 	break	6		# signal overflow
 	.set	reorder
 	$SLL	$a0,$t9
@@ -917,7 +914,8 @@ $code.=<<___;
 	$SRL	$DH,$a2,4*$BNSZ	# bits
 	sgeu	$at,$a0,$a2
 	.set	noreorder
-	bnezl	$at,.+8
+	beqz	$at,.+12
+	nop
 	$SUBU	$a0,$a2
 	.set	reorder
 
@@ -1874,6 +1872,41 @@ ___
 
 ($a_4,$a_5,$a_6,$a_7)=($b_0,$b_1,$b_2,$b_3);
 
+sub add_c2 () {
+my ($hi,$lo,$c0,$c1,$c2,
+    $warm,      # !$warm denotes first call with specific sequence of
+                # $c_[XYZ] when there is no Z-carry to accumulate yet;
+    $an,$bn     # these two are arguments for multiplication which
+                # result is used in *next* step [which is why it's
+                # commented as "forward multiplication" below];
+    )=@_;
+$code.=<<___;
+	mflo	$lo
+	mfhi	$hi
+	$ADDU	$c0,$lo
+	sltu	$at,$c0,$lo
+	 $MULTU	$an,$bn			# forward multiplication
+	$ADDU	$c0,$lo
+	$ADDU	$at,$hi
+	sltu	$lo,$c0,$lo
+	$ADDU	$c1,$at
+	$ADDU	$hi,$lo
+___
+$code.=<<___	if (!$warm);
+	sltu	$c2,$c1,$at
+	$ADDU	$c1,$hi
+	sltu	$hi,$c1,$hi
+	$ADDU	$c2,$hi
+___
+$code.=<<___	if ($warm);
+	sltu	$at,$c1,$at
+	$ADDU	$c1,$hi
+	$ADDU	$c2,$at
+	sltu	$hi,$c1,$hi
+	$ADDU	$c2,$hi
+___
+}
+
 $code.=<<___;
 
 .align	5
@@ -1922,21 +1955,10 @@ $code.=<<___;
 	sltu	$at,$c_2,$t_1
 	$ADDU	$c_3,$t_2,$at
 	$ST	$c_2,$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_2,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_1,$a_1		# mul_add_c(a[1],b[1],c3,c1,c2);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
+___
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+		$a_1,$a_1);		# mul_add_c(a[1],b[1],c3,c1,c2);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_3,$t_1
@@ -1947,67 +1969,19 @@ $code.=<<___;
 	sltu	$at,$c_1,$t_2
 	$ADDU	$c_2,$at
 	$ST	$c_3,2*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_3,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_1,$a_2		# mul_add_c2(a[1],b[2],c1,c2,c3);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_3,$at
-	 $MULTU	$a_4,$a_0		# mul_add_c2(a[4],b[0],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
+___
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+		$a_1,$a_2);		# mul_add_c2(a[1],b[2],c1,c2,c3);
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+		$a_4,$a_0);		# mul_add_c2(a[4],b[0],c2,c3,c1);
+$code.=<<___;
 	$ST	$c_1,3*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_1,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_3,$a_1		# mul_add_c2(a[3],b[1],c2,c3,c1);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_1,$at
-	$MULTU	$a_2,$a_2		# mul_add_c(a[2],b[2],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
+___
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+		$a_3,$a_1);		# mul_add_c2(a[3],b[1],c2,c3,c1);
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+		$a_2,$a_2);		# mul_add_c(a[2],b[2],c2,c3,c1);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_2,$t_1
@@ -2018,97 +1992,23 @@ $code.=<<___;
 	sltu	$at,$c_3,$t_2
 	$ADDU	$c_1,$at
 	$ST	$c_2,4*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_2,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_1,$a_4		# mul_add_c2(a[1],b[4],c3,c1,c2);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_2,$at
-	$MULTU	$a_2,$a_3		# mul_add_c2(a[2],b[3],c3,c1,c2);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	 $MULTU	$a_6,$a_0		# mul_add_c2(a[6],b[0],c1,c2,c3);
-	$ADDU	$c_2,$at
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
+___
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+		$a_1,$a_4);		# mul_add_c2(a[1],b[4],c3,c1,c2);
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+		$a_2,$a_3);		# mul_add_c2(a[2],b[3],c3,c1,c2);
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+		$a_6,$a_0);		# mul_add_c2(a[6],b[0],c1,c2,c3);
+$code.=<<___;
 	$ST	$c_3,5*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_3,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_5,$a_1		# mul_add_c2(a[5],b[1],c1,c2,c3);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_3,$at
-	$MULTU	$a_4,$a_2		# mul_add_c2(a[4],b[2],c1,c2,c3);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_3,$at
-	$MULTU	$a_3,$a_3		# mul_add_c(a[3],b[3],c1,c2,c3);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
+___
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+		$a_5,$a_1);		# mul_add_c2(a[5],b[1],c1,c2,c3);
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+		$a_4,$a_2);		# mul_add_c2(a[4],b[2],c1,c2,c3);
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+		$a_3,$a_3);		# mul_add_c(a[3],b[3],c1,c2,c3);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_1,$t_1
@@ -2119,112 +2019,25 @@ $code.=<<___;
 	sltu	$at,$c_2,$t_2
 	$ADDU	$c_3,$at
 	$ST	$c_1,6*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_1,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_1,$a_6		# mul_add_c2(a[1],b[6],c2,c3,c1);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_1,$at
-	$MULTU	$a_2,$a_5		# mul_add_c2(a[2],b[5],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_1,$at
-	$MULTU	$a_3,$a_4		# mul_add_c2(a[3],b[4],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_1,$at
-	 $MULTU	$a_7,$a_1		# mul_add_c2(a[7],b[1],c3,c1,c2);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
+___
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+		$a_1,$a_6);		# mul_add_c2(a[1],b[6],c2,c3,c1);
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+		$a_2,$a_5);		# mul_add_c2(a[2],b[5],c2,c3,c1);
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+		$a_3,$a_4);		# mul_add_c2(a[3],b[4],c2,c3,c1);
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+		$a_7,$a_1);		# mul_add_c2(a[7],b[1],c3,c1,c2);
+$code.=<<___;
 	$ST	$c_2,7*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_2,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_6,$a_2		# mul_add_c2(a[6],b[2],c3,c1,c2);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_2,$at
-	$MULTU	$a_5,$a_3		# mul_add_c2(a[5],b[3],c3,c1,c2);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_2,$at
-	$MULTU	$a_4,$a_4		# mul_add_c(a[4],b[4],c3,c1,c2);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
+___
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+		$a_6,$a_2);		# mul_add_c2(a[6],b[2],c3,c1,c2);
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+		$a_5,$a_3);		# mul_add_c2(a[5],b[3],c3,c1,c2);
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+		$a_4,$a_4);		# mul_add_c(a[4],b[4],c3,c1,c2);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_3,$t_1
@@ -2235,82 +2048,21 @@ $code.=<<___;
 	sltu	$at,$c_1,$t_2
 	$ADDU	$c_2,$at
 	$ST	$c_3,8*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_3,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_3,$a_6		# mul_add_c2(a[3],b[6],c1,c2,c3);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_3,$at
-	$MULTU	$a_4,$a_5		# mul_add_c2(a[4],b[5],c1,c2,c3);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_3,$at
-	 $MULTU	$a_7,$a_3		# mul_add_c2(a[7],b[3],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
+___
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+		$a_3,$a_6);		# mul_add_c2(a[3],b[6],c1,c2,c3);
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+		$a_4,$a_5);		# mul_add_c2(a[4],b[5],c1,c2,c3);
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+		$a_7,$a_3);		# mul_add_c2(a[7],b[3],c2,c3,c1);
+$code.=<<___;
 	$ST	$c_1,9*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_1,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_6,$a_4		# mul_add_c2(a[6],b[4],c2,c3,c1);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_1,$at
-	$MULTU	$a_5,$a_5		# mul_add_c(a[5],b[5],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
+___
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+		$a_6,$a_4);		# mul_add_c2(a[6],b[4],c2,c3,c1);
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+		$a_5,$a_5);		# mul_add_c(a[5],b[5],c2,c3,c1);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_2,$t_1
@@ -2321,52 +2073,17 @@ $code.=<<___;
 	sltu	$at,$c_3,$t_2
 	$ADDU	$c_1,$at
 	$ST	$c_2,10*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_2,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_5,$a_6		# mul_add_c2(a[5],b[6],c3,c1,c2);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_2,$at
-	 $MULTU	$a_7,$a_5		# mul_add_c2(a[7],b[5],c1,c2,c3);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
+___
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+		$a_5,$a_6);		# mul_add_c2(a[5],b[6],c3,c1,c2);
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+		$a_7,$a_5);		# mul_add_c2(a[7],b[5],c1,c2,c3);
+$code.=<<___;
 	$ST	$c_3,11*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_3,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_6,$a_6		# mul_add_c(a[6],b[6],c1,c2,c3);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
+___
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+		$a_6,$a_6);		# mul_add_c(a[6],b[6],c1,c2,c3);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_1,$t_1
@@ -2377,21 +2094,10 @@ $code.=<<___;
 	sltu	$at,$c_2,$t_2
 	$ADDU	$c_3,$at
 	$ST	$c_1,12*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_1,$t_2,$zero
-	$SLL	$t_2,1
-	 $MULTU	$a_7,$a_7		# mul_add_c(a[7],b[7],c3,c1,c2);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
+___
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+		$a_7,$a_7);		# mul_add_c(a[7],b[7],c3,c1,c2);
+$code.=<<___;
 	$ST	$c_2,13*$BNSZ($a0)
 
 	mflo	$t_1
@@ -2459,21 +2165,10 @@ $code.=<<___;
 	sltu	$at,$c_2,$t_1
 	$ADDU	$c_3,$t_2,$at
 	$ST	$c_2,$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_2,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_1,$a_1		# mul_add_c(a[1],b[1],c3,c1,c2);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
+___
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+		$a_1,$a_1);		# mul_add_c(a[1],b[1],c3,c1,c2);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_3,$t_1
@@ -2484,52 +2179,17 @@ $code.=<<___;
 	sltu	$at,$c_1,$t_2
 	$ADDU	$c_2,$at
 	$ST	$c_3,2*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_3,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_1,$a_2		# mul_add_c(a2[1],b[2],c1,c2,c3);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$at,$t_2,$zero
-	$ADDU	$c_3,$at
-	 $MULTU	$a_3,$a_1		# mul_add_c2(a[3],b[1],c2,c3,c1);
-	$SLL	$t_2,1
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_1,$t_1
-	sltu	$at,$c_1,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_2,$t_2
-	sltu	$at,$c_2,$t_2
-	$ADDU	$c_3,$at
+___
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+		$a_1,$a_2);		# mul_add_c2(a2[1],b[2],c1,c2,c3);
+	&add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+		$a_3,$a_1);		# mul_add_c2(a[3],b[1],c2,c3,c1);
+$code.=<<___;
 	$ST	$c_1,3*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_1,$t_2,$zero
-	$SLL	$t_2,1
-	$MULTU	$a_2,$a_2		# mul_add_c(a[2],b[2],c2,c3,c1);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_2,$t_1
-	sltu	$at,$c_2,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_3,$t_2
-	sltu	$at,$c_3,$t_2
-	$ADDU	$c_1,$at
+___
+	&add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+		$a_2,$a_2);		# mul_add_c(a[2],b[2],c2,c3,c1);
+$code.=<<___;
 	mflo	$t_1
 	mfhi	$t_2
 	$ADDU	$c_2,$t_1
@@ -2540,21 +2200,10 @@ $code.=<<___;
 	sltu	$at,$c_3,$t_2
 	$ADDU	$c_1,$at
 	$ST	$c_2,4*$BNSZ($a0)
-
-	mflo	$t_1
-	mfhi	$t_2
-	slt	$c_2,$t_2,$zero
-	$SLL	$t_2,1
-	 $MULTU	$a_3,$a_3		# mul_add_c(a[3],b[3],c1,c2,c3);
-	slt	$a2,$t_1,$zero
-	$ADDU	$t_2,$a2
-	$SLL	$t_1,1
-	$ADDU	$c_3,$t_1
-	sltu	$at,$c_3,$t_1
-	$ADDU	$t_2,$at
-	$ADDU	$c_1,$t_2
-	sltu	$at,$c_1,$t_2
-	$ADDU	$c_2,$at
+___
+	&add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+		$a_3,$a_3);		# mul_add_c(a[3],b[3],c1,c2,c3);
+$code.=<<___;
 	$ST	$c_3,5*$BNSZ($a0)
 
 	mflo	$t_1
diff --git a/lib/libssl/src/crypto/bn/asm/x86_64-gcc.c b/lib/libssl/src/crypto/bn/asm/x86_64-gcc.c
index c9a2b6be734..9deffa71f1a 100644
--- a/lib/libssl/src/crypto/bn/asm/x86_64-gcc.c
+++ b/lib/libssl/src/crypto/bn/asm/x86_64-gcc.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: x86_64-gcc.c,v 1.4 2014/10/28 07:35:58 jsg Exp $ */
+/* $OpenBSD: x86_64-gcc.c,v 1.5 2015/02/25 15:39:49 bcook Exp $ */
 #include "../bn_lcl.h"
 #if !(defined(__GNUC__) && __GNUC__>=2)
 # include "../bn_asm.c"	/* kind of dirty hack for Sun Studio */
@@ -270,77 +270,76 @@ BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
 /* sqr_add_c(a,i,c0,c1,c2)  -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
 /* sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number c=(c2,c1,c0) */
 
+/*
+ * Keep in mind that carrying into high part of multiplication result
+ * can not overflow, because it cannot be all-ones.
+ */
 #if 0
 /* original macros are kept for reference purposes */
-#define mul_add_c(a,b,c0,c1,c2) {	\
-	BN_ULONG ta=(a),tb=(b);		\
-	t1 = ta * tb;			\
-	t2 = BN_UMULT_HIGH(ta,tb);	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
-
-#define mul_add_c2(a,b,c0,c1,c2) {	\
-	BN_ULONG ta=(a),tb=(b),t0;	\
-	t1 = BN_UMULT_HIGH(ta,tb);	\
-	t0 = ta * tb;			\
-	t2 = t1+t1; c2 += (t2<t1)?1:0;	\
-	t1 = t0+t0; t2 += (t1<t0)?1:0;	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
+#define mul_add_c(a,b,c0,c1,c2)		do {	\
+	BN_ULONG ta = (a), tb = (b);		\
+	BN_ULONG lo, hi;			\
+	BN_UMULT_LOHI(lo,hi,ta,tb);		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2)	do {	\
+	BN_ULONG ta = (a), tb = (b);		\
+	BN_ULONG lo, hi, tt;			\
+	BN_UMULT_LOHI(lo,hi,ta,tb);		\
+	c0 += lo; tt = hi+((c0<lo)?1:0);	\
+	c1 += tt; c2 += (c1<tt)?1:0;		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2)		do {	\
+	BN_ULONG ta = (a)[i];			\
+	BN_ULONG lo, hi;			\
+	BN_UMULT_LOHI(lo,hi,ta,ta);		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
 #else
 #define mul_add_c(a,b,c0,c1,c2)	do {	\
+	BN_ULONG t1,t2;			\
 	asm ("mulq %3"			\
 		: "=a"(t1),"=d"(t2)	\
 		: "a"(a),"m"(b)		\
 		: "cc");		\
-	asm ("addq %2,%0; adcq %3,%1"	\
-		: "+r"(c0),"+d"(t2)	\
-		: "a"(t1),"g"(0)	\
-		: "cc");		\
-	asm ("addq %2,%0; adcq %3,%1"	\
-		: "+r"(c1),"+r"(c2)	\
-		: "d"(t2),"g"(0)	\
-		: "cc");		\
+	asm ("addq %3,%0; adcq %4,%1; adcq %5,%2"	\
+		: "+r"(c0),"+r"(c1),"+r"(c2)		\
+		: "r"(t1),"r"(t2),"g"(0)		\
+		: "cc");				\
 	} while (0)
 
 #define sqr_add_c(a,i,c0,c1,c2)	do {	\
+	BN_ULONG t1,t2;			\
 	asm ("mulq %2"			\
 		: "=a"(t1),"=d"(t2)	\
 		: "a"(a[i])		\
 		: "cc");		\
-	asm ("addq %2,%0; adcq %3,%1"	\
-		: "+r"(c0),"+d"(t2)	\
-		: "a"(t1),"g"(0)	\
-		: "cc");		\
-	asm ("addq %2,%0; adcq %3,%1"	\
-		: "+r"(c1),"+r"(c2)	\
-		: "d"(t2),"g"(0)	\
-		: "cc");		\
+	asm ("addq %3,%0; adcq %4,%1; adcq %5,%2"	\
+		: "+r"(c0),"+r"(c1),"+r"(c2)		\
+		: "r"(t1),"r"(t2),"g"(0)		\
+		: "cc");				\
 	} while (0)
 
 #define mul_add_c2(a,b,c0,c1,c2) do {	\
+	BN_ULONG t1,t2;			\
 	asm ("mulq %3"			\
 		: "=a"(t1),"=d"(t2)	\
 		: "a"(a),"m"(b)		\
 		: "cc");		\
-	asm ("addq %0,%0; adcq %2,%1"	\
-		: "+d"(t2),"+r"(c2)	\
-		: "g"(0)		\
-		: "cc");		\
-	asm ("addq %0,%0; adcq %2,%1"	\
-		: "+a"(t1),"+d"(t2)	\
-		: "g"(0)		\
-		: "cc");		\
-	asm ("addq %2,%0; adcq %3,%1"	\
-		: "+r"(c0),"+d"(t2)	\
-		: "a"(t1),"g"(0)	\
-		: "cc");		\
-	asm ("addq %2,%0; adcq %3,%1"	\
-		: "+r"(c1),"+r"(c2)	\
-		: "d"(t2),"g"(0)	\
-		: "cc");		\
+	asm ("addq %3,%0; adcq %4,%1; adcq %5,%2"	\
+		: "+r"(c0),"+r"(c1),"+r"(c2)		\
+		: "r"(t1),"r"(t2),"g"(0)		\
+		: "cc");				\
+	asm ("addq %3,%0; adcq %4,%1; adcq %5,%2"	\
+		: "+r"(c0),"+r"(c1),"+r"(c2)		\
+		: "r"(t1),"r"(t2),"g"(0)		\
+		: "cc");				\
 	} while (0)
 #endif
 
@@ -349,7 +348,6 @@ BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
 
 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 	{
-	BN_ULONG t1,t2;
 	BN_ULONG c1,c2,c3;
 
 	c1=0;
@@ -453,7 +451,6 @@ void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 
 void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 	{
-	BN_ULONG t1,t2;
 	BN_ULONG c1,c2,c3;
 
 	c1=0;
@@ -493,7 +490,6 @@ void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 
 void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
 	{
-	BN_ULONG t1,t2;
 	BN_ULONG c1,c2,c3;
 
 	c1=0;
@@ -569,7 +565,6 @@ void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
 
 void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
 	{
-	BN_ULONG t1,t2;
 	BN_ULONG c1,c2,c3;
 
 	c1=0;
diff --git a/lib/libssl/src/crypto/bn/bn_asm.c b/lib/libssl/src/crypto/bn/bn_asm.c
index c6efd2513ac..49f0ba5d7b5 100644
--- a/lib/libssl/src/crypto/bn/bn_asm.c
+++ b/lib/libssl/src/crypto/bn/bn_asm.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: bn_asm.c,v 1.13 2014/07/11 08:44:47 jsing Exp $ */
+/* $OpenBSD: bn_asm.c,v 1.14 2015/02/25 15:39:49 bcook Exp $ */
 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
  * All rights reserved.
  *
@@ -495,116 +495,143 @@ bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
 /* sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number c=(c2,c1,c0) */
 
 #ifdef BN_LLONG
-#define mul_add_c(a,b,c0,c1,c2) \
-	t=(BN_ULLONG)a*b; \
-	t1=(BN_ULONG)Lw(t); \
-	t2=(BN_ULONG)Hw(t); \
-	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
-	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define mul_add_c2(a,b,c0,c1,c2) \
-	t=(BN_ULLONG)a*b; \
-	tt=(t+t)&BN_MASK; \
-	if (tt < t) c2++; \
-	t1=(BN_ULONG)Lw(tt); \
-	t2=(BN_ULONG)Hw(tt); \
-	c0=(c0+t1)&BN_MASK2;  \
-	if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
-	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define sqr_add_c(a,i,c0,c1,c2) \
-	t=(BN_ULLONG)a[i]*a[i]; \
-	t1=(BN_ULONG)Lw(t); \
-	t2=(BN_ULONG)Hw(t); \
-	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
-	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
+/*
+ * Keep in mind that additions to multiplication result can not
+ * overflow, because its high half cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2)		do {	\
+	BN_ULONG hi;				\
+	BN_ULLONG t = (BN_ULLONG)(a)*(b);	\
+	t += c0;		/* no carry */	\
+	c0 = (BN_ULONG)Lw(t);			\
+	hi = (BN_ULONG)Hw(t);			\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	} while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2)	do {	\
+	BN_ULONG hi;				\
+	BN_ULLONG t = (BN_ULLONG)(a)*(b);	\
+	BN_ULLONG tt = t+c0;	/* no carry */	\
+	c0 = (BN_ULONG)Lw(tt);			\
+	hi = (BN_ULONG)Hw(tt);			\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	t += c0;		/* no carry */	\
+	c0 = (BN_ULONG)Lw(t);			\
+	hi = (BN_ULONG)Hw(t);			\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	} while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2)		do {	\
+	BN_ULONG hi;				\
+	BN_ULLONG t = (BN_ULLONG)a[i]*a[i];	\
+	t += c0;		/* no carry */	\
+	c0 = (BN_ULONG)Lw(t);			\
+	hi = (BN_ULONG)Hw(t);			\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	} while(0)
 
 #define sqr_add_c2(a,i,j,c0,c1,c2) \
 	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
 
 #elif defined(BN_UMULT_LOHI)
-
-#define mul_add_c(a,b,c0,c1,c2)	{	\
-	BN_ULONG ta=(a),tb=(b);		\
-	BN_UMULT_LOHI(t1,t2,ta,tb);	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
-
-#define mul_add_c2(a,b,c0,c1,c2) {	\
-	BN_ULONG ta=(a),tb=(b),t0;	\
-	BN_UMULT_LOHI(t0,t1,ta,tb);	\
-	t2 = t1+t1; c2 += (t2<t1)?1:0;	\
-	t1 = t0+t0; t2 += (t1<t0)?1:0;	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
-
-#define sqr_add_c(a,i,c0,c1,c2)	{	\
-	BN_ULONG ta=(a)[i];		\
-	BN_UMULT_LOHI(t1,t2,ta,ta);	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2)		do {	\
+	BN_ULONG ta = (a), tb = (b);		\
+	BN_ULONG lo, hi;			\
+	BN_UMULT_LOHI(lo,hi,ta,tb);		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2)	do {	\
+	BN_ULONG ta = (a), tb = (b);		\
+	BN_ULONG lo, hi, tt;			\
+	BN_UMULT_LOHI(lo,hi,ta,tb);		\
+	c0 += lo; tt = hi+((c0<lo)?1:0);	\
+	c1 += tt; c2 += (c1<tt)?1:0;		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2)		do {	\
+	BN_ULONG ta = (a)[i];			\
+	BN_ULONG lo, hi;			\
+	BN_UMULT_LOHI(lo,hi,ta,ta);		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
 
 #define sqr_add_c2(a,i,j,c0,c1,c2)	\
 	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
 
 #elif defined(BN_UMULT_HIGH)
-
-#define mul_add_c(a,b,c0,c1,c2)	{	\
-	BN_ULONG ta=(a),tb=(b);		\
-	t1 = ta * tb;			\
-	t2 = BN_UMULT_HIGH(ta,tb);	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
-
-#define mul_add_c2(a,b,c0,c1,c2) {	\
-	BN_ULONG ta=(a),tb=(b),t0;	\
-	t1 = BN_UMULT_HIGH(ta,tb);	\
-	t0 = ta * tb;			\
-	t2 = t1+t1; c2 += (t2<t1)?1:0;	\
-	t1 = t0+t0; t2 += (t1<t0)?1:0;	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
-
-#define sqr_add_c(a,i,c0,c1,c2)	{	\
-	BN_ULONG ta=(a)[i];		\
-	t1 = ta * ta;			\
-	t2 = BN_UMULT_HIGH(ta,ta);	\
-	c0 += t1; t2 += (c0<t1)?1:0;	\
-	c1 += t2; c2 += (c1<t2)?1:0;	\
-	}
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2)		do {	\
+	BN_ULONG ta = (a), tb = (b);		\
+	BN_ULONG lo = ta * tb;			\
+	BN_ULONG hi = BN_UMULT_HIGH(ta,tb);	\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2)	do {	\
+	BN_ULONG ta = (a), tb = (b), tt;	\
+	BN_ULONG lo = ta * tb;			\
+	BN_ULONG hi = BN_UMULT_HIGH(ta,tb);	\
+	c0 += lo; tt = hi + ((c0<lo)?1:0);	\
+	c1 += tt; c2 += (c1<tt)?1:0;		\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2)		do {	\
+	BN_ULONG ta = (a)[i];			\
+	BN_ULONG lo = ta * ta;			\
+	BN_ULONG hi = BN_UMULT_HIGH(ta,ta);	\
+	c0 += lo; hi += (c0<lo)?1:0;		\
+	c1 += hi; c2 += (c1<hi)?1:0;		\
+	} while(0)
 
 #define sqr_add_c2(a,i,j,c0,c1,c2)	\
 	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
 
 #else /* !BN_LLONG */
-#define mul_add_c(a,b,c0,c1,c2) \
-	t1=LBITS(a); t2=HBITS(a); \
-	bl=LBITS(b); bh=HBITS(b); \
-	mul64(t1,t2,bl,bh); \
-	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
-	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define mul_add_c2(a,b,c0,c1,c2) \
-	t1=LBITS(a); t2=HBITS(a); \
-	bl=LBITS(b); bh=HBITS(b); \
-	mul64(t1,t2,bl,bh); \
-	if (t2 & BN_TBIT) c2++; \
-	t2=(t2+t2)&BN_MASK2; \
-	if (t1 & BN_TBIT) t2++; \
-	t1=(t1+t1)&BN_MASK2; \
-	c0=(c0+t1)&BN_MASK2;  \
-	if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
-	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
-
-#define sqr_add_c(a,i,c0,c1,c2) \
-	sqr64(t1,t2,(a)[i]); \
-	c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
-	c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+#define mul_add_c(a,b,c0,c1,c2)		do {	\
+	BN_ULONG lo = LBITS(a), hi = HBITS(a);	\
+	BN_ULONG bl = LBITS(b), bh = HBITS(b);	\
+	mul64(lo,hi,bl,bh);			\
+	c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++;	\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	} while(0)
+
+#define mul_add_c2(a,b,c0,c1,c2)	do {	\
+	BN_ULONG tt;				\
+	BN_ULONG lo = LBITS(a), hi = HBITS(a);	\
+	BN_ULONG bl = LBITS(b), bh = HBITS(b);	\
+	mul64(lo,hi,bl,bh);			\
+	tt = hi;				\
+	c0 = (c0+lo)&BN_MASK2; if (c0<lo) tt++;	\
+	c1 = (c1+tt)&BN_MASK2; if (c1<tt) c2++;	\
+	c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++;	\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	} while(0)
+
+#define sqr_add_c(a,i,c0,c1,c2)		do {	\
+	BN_ULONG lo, hi;			\
+	sqr64(lo,hi,(a)[i]);			\
+	c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++;	\
+	c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++;	\
+	} while(0)
 
 #define sqr_add_c2(a,i,j,c0,c1,c2) \
 	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
@@ -613,12 +640,6 @@ bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
 void
 bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 {
-#ifdef BN_LLONG
-	BN_ULLONG t;
-#elif !defined(BN_UMULT_LOHI) && !defined(BN_UMULT_HIGH)
-	BN_ULONG bl, bh;
-#endif
-	BN_ULONG t1, t2;
 	BN_ULONG c1, c2, c3;
 
 	c1 = 0;
@@ -723,12 +744,6 @@ bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 void
 bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 {
-#ifdef BN_LLONG
-	BN_ULLONG t;
-#elif !defined(BN_UMULT_LOHI) && !defined(BN_UMULT_HIGH)
-	BN_ULONG bl, bh;
-#endif
-	BN_ULONG t1, t2;
 	BN_ULONG c1, c2, c3;
 
 	c1 = 0;
@@ -769,12 +784,6 @@ bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 void
 bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
 {
-#ifdef BN_LLONG
-	BN_ULLONG t, tt;
-#elif !defined(BN_UMULT_LOHI) && !defined(BN_UMULT_HIGH)
-	BN_ULONG bl, bh;
-#endif
-	BN_ULONG t1, t2;
 	BN_ULONG c1, c2, c3;
 
 	c1 = 0;
@@ -851,12 +860,6 @@ bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
 void
 bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
 {
-#ifdef BN_LLONG
-	BN_ULLONG t, tt;
-#elif !defined(BN_UMULT_LOHI) && !defined(BN_UMULT_HIGH)
-	BN_ULONG bl, bh;
-#endif
-	BN_ULONG t1, t2;
 	BN_ULONG c1, c2, c3;
 
 	c1 = 0;