X-Git-Url: https://git.openssl.org/gitweb/?p=openssl.git;a=blobdiff_plain;f=crypto%2Fbn%2Fbn_asm.c;h=741cf813dbe9e5905b815d2d55ac1868bb723e91;hp=286a0f1e74b79967a66d30fffcee3a4c7ed17226;hb=682b112abc7489a0e24039b9516457f13bb91309;hpb=cea538154bfb543b5fe1d2556a779cccd2572a5e

diff --git a/crypto/bn/bn_asm.c b/crypto/bn/bn_asm.c
index 286a0f1e74..741cf813db 100644
--- a/crypto/bn/bn_asm.c
+++ b/crypto/bn/bn_asm.c
@@ -56,169 +56,174 @@
  * [including the GNU Public Licence.]
  */
 
+#ifndef BN_DEBUG
+# undef NDEBUG /* avoid conflicting definitions */
+# define NDEBUG
+#endif
+
 #include <stdio.h>
+#include <assert.h>
 #include "cryptlib.h"
 #include "bn_lcl.h"
 
-#ifdef BN_LLONG 
+#if defined(BN_LLONG) || defined(BN_UMULT_HIGH)
 
-BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
 	{
 	BN_ULONG c1=0;
 
-	bn_check_num(num);
+	assert(num >= 0);
 	if (num <= 0) return(c1);
 
-	for (;;)
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (num&~3)
 		{
 		mul_add(rp[0],ap[0],w,c1);
-		if (--num == 0) break;
 		mul_add(rp[1],ap[1],w,c1);
-		if (--num == 0) break;
 		mul_add(rp[2],ap[2],w,c1);
-		if (--num == 0) break;
 		mul_add(rp[3],ap[3],w,c1);
-		if (--num == 0) break;
-		ap+=4;
-		rp+=4;
+		ap+=4; rp+=4; num-=4;
+		}
+#endif
+	while (num)
+		{
+		mul_add(rp[0],ap[0],w,c1);
+		ap++; rp++; num--;
 		}
 	
 	return(c1);
 	} 
 
-BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
 	{
 	BN_ULONG c1=0;
 
-	bn_check_num(num);
+	assert(num >= 0);
 	if (num <= 0) return(c1);
 
-	/* for (;;) */
-	while (1) /* circumvent egcs-1.1.2 bug */
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (num&~3)
 		{
 		mul(rp[0],ap[0],w,c1);
-		if (--num == 0) break;
 		mul(rp[1],ap[1],w,c1);
-		if (--num == 0) break;
 		mul(rp[2],ap[2],w,c1);
-		if (--num == 0) break;
 		mul(rp[3],ap[3],w,c1);
-		if (--num == 0) break;
-		ap+=4;
-		rp+=4;
+		ap+=4; rp+=4; num-=4;
+		}
+#endif
+	while (num)
+		{
+		mul(rp[0],ap[0],w,c1);
+		ap++; rp++; num--;
 		}
 	return(c1);
 	} 
 
-void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
+void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
         {
-	bn_check_num(n);
+	assert(n >= 0);
 	if (n <= 0) return;
-	for (;;)
-		{
-		BN_ULLONG t;
-
-		t=(BN_ULLONG)(a[0])*(a[0]);
-		r[0]=Lw(t); r[1]=Hw(t);
-		if (--n == 0) break;
-
-		t=(BN_ULLONG)(a[1])*(a[1]);
-		r[2]=Lw(t); r[3]=Hw(t);
-		if (--n == 0) break;
 
-		t=(BN_ULLONG)(a[2])*(a[2]);
-		r[4]=Lw(t); r[5]=Hw(t);
-		if (--n == 0) break;
-
-		t=(BN_ULLONG)(a[3])*(a[3]);
-		r[6]=Lw(t); r[7]=Hw(t);
-		if (--n == 0) break;
-
-		a+=4;
-		r+=8;
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (n&~3)
+		{
+		sqr(r[0],r[1],a[0]);
+		sqr(r[2],r[3],a[1]);
+		sqr(r[4],r[5],a[2]);
+		sqr(r[6],r[7],a[3]);
+		a+=4; r+=8; n-=4;
+		}
+#endif
+	while (n)
+		{
+		sqr(r[0],r[1],a[0]);
+		a++; r+=2; n--;
 		}
 	}
 
-#else
+#else /* !(defined(BN_LLONG) || defined(BN_UMULT_HIGH)) */
 
-BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
 	{
 	BN_ULONG c=0;
 	BN_ULONG bl,bh;
 
-	bn_check_num(num);
+	assert(num >= 0);
 	if (num <= 0) return((BN_ULONG)0);
 
 	bl=LBITS(w);
 	bh=HBITS(w);
 
-	for (;;)
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (num&~3)
 		{
 		mul_add(rp[0],ap[0],bl,bh,c);
-		if (--num == 0) break;
 		mul_add(rp[1],ap[1],bl,bh,c);
-		if (--num == 0) break;
 		mul_add(rp[2],ap[2],bl,bh,c);
-		if (--num == 0) break;
 		mul_add(rp[3],ap[3],bl,bh,c);
-		if (--num == 0) break;
-		ap+=4;
-		rp+=4;
+		ap+=4; rp+=4; num-=4;
+		}
+#endif
+	while (num)
+		{
+		mul_add(rp[0],ap[0],bl,bh,c);
+		ap++; rp++; num--;
 		}
 	return(c);
 	} 
 
-BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
 	{
 	BN_ULONG carry=0;
 	BN_ULONG bl,bh;
 
-	bn_check_num(num);
+	assert(num >= 0);
 	if (num <= 0) return((BN_ULONG)0);
 
 	bl=LBITS(w);
 	bh=HBITS(w);
 
-	for (;;)
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (num&~3)
 		{
 		mul(rp[0],ap[0],bl,bh,carry);
-		if (--num == 0) break;
 		mul(rp[1],ap[1],bl,bh,carry);
-		if (--num == 0) break;
 		mul(rp[2],ap[2],bl,bh,carry);
-		if (--num == 0) break;
 		mul(rp[3],ap[3],bl,bh,carry);
-		if (--num == 0) break;
-		ap+=4;
-		rp+=4;
+		ap+=4; rp+=4; num-=4;
+		}
+#endif
+	while (num)
+		{
+		mul(rp[0],ap[0],bl,bh,carry);
+		ap++; rp++; num--;
 		}
 	return(carry);
 	} 
 
-void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
+void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
         {
-	bn_check_num(n);
+	assert(n >= 0);
 	if (n <= 0) return;
-	for (;;)
+
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (n&~3)
 		{
 		sqr64(r[0],r[1],a[0]);
-		if (--n == 0) break;
-
 		sqr64(r[2],r[3],a[1]);
-		if (--n == 0) break;
-
 		sqr64(r[4],r[5],a[2]);
-		if (--n == 0) break;
-
 		sqr64(r[6],r[7],a[3]);
-		if (--n == 0) break;
-
-		a+=4;
-		r+=8;
+		a+=4; r+=8; n-=4;
+		}
+#endif
+	while (n)
+		{
+		sqr64(r[0],r[1],a[0]);
+		a++; r+=2; n--;
 		}
 	}
 
-#endif
+#endif /* !(defined(BN_LLONG) || defined(BN_UMULT_HIGH)) */
 
 #if defined(BN_LLONG) && defined(BN_DIV2W)
 
@@ -229,7 +234,7 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 
 #else
 
-/* Divide h-l by d and return the result. */
+/* Divide h,l by d and return the result. */
 /* I need to test this some more :-( */
 BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 	{
@@ -239,13 +244,8 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 	if (d == 0) return(BN_MASK2);
 
 	i=BN_num_bits_word(d);
-	if ((i != BN_BITS2) && (h > (BN_ULONG)1<<i))
-		{
-#if !defined(NO_STDIO) && !defined(WIN16)
-		fprintf(stderr,"Division would overflow (%d)\n",i);
-#endif
-		abort();
-		}
+	assert((i == BN_BITS2) || (h <= (BN_ULONG)1<<i));
+
 	i=BN_BITS2-i;
 	if (h >= d) h-=d;
 
@@ -264,18 +264,20 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 		else
 			q=h/dh;
 
+		th=q*dh;
+		tl=dl*q;
 		for (;;)
 			{
-			t=(h-(th=q*dh));
-			tl=BN_MASK2;
+			t=h-th;
 			if ((t&BN_MASK2h) ||
-				((tl=dl*q) <= (
+				((tl) <= (
 					(t<<BN_BITS4)|
 					((l&BN_MASK2h)>>BN_BITS4))))
 				break;
 			q--;
+			th-=dh;
+			tl-=dl;
 			}
-		if (tl==BN_MASK2) tl=q*dl;
 		t=(tl>>BN_BITS4);
 		tl=(tl<<BN_BITS4)&BN_MASK2h;
 		th+=t;
@@ -298,54 +300,54 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
 	ret|=q;
 	return(ret);
 	}
-#endif
+#endif /* !defined(BN_LLONG) && defined(BN_DIV2W) */
 
 #ifdef BN_LLONG
-BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
         {
 	BN_ULLONG ll=0;
 
-	bn_check_num(n);
+	assert(n >= 0);
 	if (n <= 0) return((BN_ULONG)0);
 
-	for (;;)
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (n&~3)
 		{
 		ll+=(BN_ULLONG)a[0]+b[0];
 		r[0]=(BN_ULONG)ll&BN_MASK2;
 		ll>>=BN_BITS2;
-		if (--n <= 0) break;
-
 		ll+=(BN_ULLONG)a[1]+b[1];
 		r[1]=(BN_ULONG)ll&BN_MASK2;
 		ll>>=BN_BITS2;
-		if (--n <= 0) break;
-
 		ll+=(BN_ULLONG)a[2]+b[2];
 		r[2]=(BN_ULONG)ll&BN_MASK2;
 		ll>>=BN_BITS2;
-		if (--n <= 0) break;
-
 		ll+=(BN_ULLONG)a[3]+b[3];
 		r[3]=(BN_ULONG)ll&BN_MASK2;
 		ll>>=BN_BITS2;
-		if (--n <= 0) break;
-
-		a+=4;
-		b+=4;
-		r+=4;
+		a+=4; b+=4; r+=4; n-=4;
+		}
+#endif
+	while (n)
+		{
+		ll+=(BN_ULLONG)a[0]+b[0];
+		r[0]=(BN_ULONG)ll&BN_MASK2;
+		ll>>=BN_BITS2;
+		a++; b++; r++; n--;
 		}
 	return((BN_ULONG)ll);
 	}
-#else
-BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+#else /* !BN_LLONG */
+BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
         {
 	BN_ULONG c,l,t;
 
-	bn_check_num(n);
+	assert(n >= 0);
 	if (n <= 0) return((BN_ULONG)0);
 
 	c=0;
-	for (;;)
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (n&~3)
 		{
 		t=a[0];
 		t=(t+c)&BN_MASK2;
@@ -353,84 +355,89 @@ BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
 		l=(t+b[0])&BN_MASK2;
 		c+=(l < t);
 		r[0]=l;
-		if (--n <= 0) break;
-
 		t=a[1];
 		t=(t+c)&BN_MASK2;
 		c=(t < c);
 		l=(t+b[1])&BN_MASK2;
 		c+=(l < t);
 		r[1]=l;
-		if (--n <= 0) break;
-
 		t=a[2];
 		t=(t+c)&BN_MASK2;
 		c=(t < c);
 		l=(t+b[2])&BN_MASK2;
 		c+=(l < t);
 		r[2]=l;
-		if (--n <= 0) break;
-
 		t=a[3];
 		t=(t+c)&BN_MASK2;
 		c=(t < c);
 		l=(t+b[3])&BN_MASK2;
 		c+=(l < t);
 		r[3]=l;
-		if (--n <= 0) break;
-
-		a+=4;
-		b+=4;
-		r+=4;
+		a+=4; b+=4; r+=4; n-=4;
+		}
+#endif
+	while(n)
+		{
+		t=a[0];
+		t=(t+c)&BN_MASK2;
+		c=(t < c);
+		l=(t+b[0])&BN_MASK2;
+		c+=(l < t);
+		r[0]=l;
+		a++; b++; r++; n--;
 		}
 	return((BN_ULONG)c);
 	}
-#endif
+#endif /* !BN_LLONG */
 
-BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+BN_ULONG bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, int n)
         {
 	BN_ULONG t1,t2;
 	int c=0;
 
-	bn_check_num(n);
+	assert(n >= 0);
 	if (n <= 0) return((BN_ULONG)0);
 
-	for (;;)
+#ifndef OPENSSL_SMALL_FOOTPRINT
+	while (n&~3)
 		{
 		t1=a[0]; t2=b[0];
 		r[0]=(t1-t2-c)&BN_MASK2;
 		if (t1 != t2) c=(t1 < t2);
-		if (--n <= 0) break;
-
 		t1=a[1]; t2=b[1];
 		r[1]=(t1-t2-c)&BN_MASK2;
 		if (t1 != t2) c=(t1 < t2);
-		if (--n <= 0) break;
-
 		t1=a[2]; t2=b[2];
 		r[2]=(t1-t2-c)&BN_MASK2;
 		if (t1 != t2) c=(t1 < t2);
-		if (--n <= 0) break;
-
 		t1=a[3]; t2=b[3];
 		r[3]=(t1-t2-c)&BN_MASK2;
 		if (t1 != t2) c=(t1 < t2);
-		if (--n <= 0) break;
-
-		a+=4;
-		b+=4;
-		r+=4;
+		a+=4; b+=4; r+=4; n-=4;
+		}
+#endif
+	while (n)
+		{
+		t1=a[0]; t2=b[0];
+		r[0]=(t1-t2-c)&BN_MASK2;
+		if (t1 != t2) c=(t1 < t2);
+		a++; b++; r++; n--;
 		}
 	return(c);
 	}
 
-#ifdef BN_MUL_COMBA
+#if defined(BN_MUL_COMBA) && !defined(OPENSSL_SMALL_FOOTPRINT)
 
 #undef bn_mul_comba8
 #undef bn_mul_comba4
 #undef bn_sqr_comba8
 #undef bn_sqr_comba4
 
+/* mul_add_c(a,b,c0,c1,c2)  -- c+=a*b for three word number c=(c2,c1,c0) */
+/* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
+/* 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) */
+
 #ifdef BN_LLONG
 #define mul_add_c(a,b,c0,c1,c2) \
 	t=(BN_ULLONG)a*b; \
@@ -458,7 +465,67 @@ BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
 
 #define sqr_add_c2(a,i,j,c0,c1,c2) \
 	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
-#else
+
+#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;	\
+	}
+
+#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;	\
+	}
+
+#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); \
@@ -485,7 +552,7 @@ BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
 
 #define sqr_add_c2(a,i,j,c0,c1,c2) \
 	mul_add_c2((a)[i],(a)[j],c0,c1,c2)
-#endif
+#endif /* !BN_LLONG */
 
 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 	{
@@ -641,7 +708,7 @@ void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 	r[7]=c2;
 	}
 
-void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
 	{
 #ifdef BN_LLONG
 	BN_ULLONG t,tt;
@@ -722,7 +789,7 @@ void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
 	r[15]=c1;
 	}
 
-void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
 	{
 #ifdef BN_LLONG
 	BN_ULLONG t,tt;
@@ -760,18 +827,95 @@ void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
 	r[6]=c1;
 	r[7]=c2;
 	}
+
+#ifdef OPENSSL_BN_ASM_MONT
+/*
+ * This is essentially reference implementation, which may or may not
+ * result in performance improvement. E.g. on IA-32 this does give 40%
+ * faster rsa1024 private key operations and 10% faster rsa4096 ones,
+ * while on AMD64 it improves rsa1024 sign only by 10% and *worsens*
+ * rsa4096 sign by 15%. Once again, it's a reference implementation,
+ * one to be used as start-point for platform-specific assembler.
+ */
+void bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,BN_ULONG n0, int num)
+	{
+	BN_ULONG c0,c1,ml,*tp;
+#ifdef mul64
+	BN_ULONG mh;
+#endif
+	volatile BN_ULONG *vp;
+	int i=0,j;
+
+	vp = tp = alloca((num+2)*sizeof(BN_ULONG));
+
+	tp[num]   = bn_mul_words(tp,ap,num,bp[0]);
+	tp[num+1] = 0;
+	goto enter;
+
+	for(i=0;i<num;i++)
+		{
+		c0 = bn_mul_add_words(tp,ap,num,bp[i]);
+		c1 = (tp[num] + c0)&BN_MASK2;
+		tp[num]   = c1;
+		tp[num+1] = (c1<c0?1:0);
+	enter:
+		c1  = tp[0];
+		ml = (c1*n0)&BN_MASK2;
+		c0 = 0;
+#ifdef mul64
+		mh = HBITS(ml);
+		ml = LBITS(ml);
+		mul_add(c1,np[0],ml,mh,c0);
 #else
+		mul_add(c1,ml,np[0],c0);
+#endif
+		for(j=1;j<num;j++)
+			{
+			c1 = tp[j];
+#ifdef mul64
+			mul_add(c1,np[j],ml,mh,c0);
+#else
+			mul_add(c1,ml,np[j],c0);
+#endif
+			tp[j-1] = c1&BN_MASK2;
+			}
+		c1        = (tp[num] + c0)&BN_MASK2;
+		tp[num-1] = c1;
+		tp[num]   = tp[num+1] + (c1<c0?1:0);
+		}
+
+	if (tp[num]!=0 || tp[num-1]>=np[num-1])
+		{
+		c0 = bn_sub_words(rp,tp,np,num);
+		if (tp[num]!=0 || c0==0)
+			{
+			for(i=0;i<num+2;i++)	vp[i] = 0;
+			return;
+			}
+		}
+	for(i=0;i<num;i++)	rp[i] = tp[i],	vp[i] = 0;
+	vp[num]   = 0;
+	vp[num+1] = 0;
+	}
+
+void bn_sqr_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *np,BN_ULONG n0, int num)
+	{
+	bn_mul_mont(rp,ap,ap,np,n0,num);
+	}
+#endif /* OPENSSL_BN_ASM_MONT */
+
+#else /* !BN_MUL_COMBA */
 
 /* hmm... is it faster just to do a multiply? */
 #undef bn_sqr_comba4
-void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
 	{
 	BN_ULONG t[8];
 	bn_sqr_normal(r,a,4,t);
 	}
 
 #undef bn_sqr_comba8
-void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
 	{
 	BN_ULONG t[16];
 	bn_sqr_normal(r,a,8,t);
@@ -797,4 +941,49 @@ void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
 	r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
 	}
 
-#endif /* BN_COMBA */
+#ifdef OPENSSL_BN_ASM_MONT
+void bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,BN_ULONG n0, int num)
+	{
+	BN_ULONG c0,c1,*tp;
+	volatile BN_ULONG *vp;
+	int i=0,j;
+
+	vp = tp = alloca((num+2)*sizeof(BN_ULONG));
+
+	for(i=0;i<=num;i++)	tp[i]=0;
+
+	for(i=0;i<num;i++)
+		{
+		c0         = bn_mul_add_words(tp,ap,num,bp[i]);
+		c1         = tp[num] + c0;
+		tp[num]    = c1;
+		tp[num+1]  = (c1<c0?1:0);
+
+		c0         = bn_mul_add_words(tp,np,num,tp[0]*n0);
+		c1         = tp[num] + c0;
+		tp[num]    = c1;
+		tp[num+1] += (c1<c0?1:0);
+		for(j=0;j<=num;j++)	tp[j]=tp[j+1];
+		}
+
+	if (tp[num]!=0 || tp[num-1]>=np[num-1])
+		{
+		c0 = bn_sub_words(rp,tp,np,num);
+		if (tp[num]!=0 || c0==0)
+			{
+			for(i=0;i<num+2;i++)	vp[i] = 0;
+			return;
+			}
+		}
+	for(i=0;i<num;i++)	rp[i] = tp[i],	vp[i] = 0;
+	vp[num]   = 0;
+	vp[num+1] = 0;
+	}
+
+void bn_sqr_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *np,BN_ULONG n0, int num)
+	{
+	bn_mul_mont(rp,ap,ap,np,n0,num);
+	}
+#endif /* OPENSSL_BN_ASM_MONT */
+
+#endif /* !BN_MUL_COMBA */