Use basename of objxref.pl not whole path in generated header.
[oweals/openssl.git] / crypto / bn / bn_nist.c
index c3356b7cf900ed2a1e3e1ea616f0786242f7a998..2ca5b0139111f76352b3d2c08ba6232056efb942 100644 (file)
@@ -73,12 +73,22 @@ static const BN_ULONG _nist_p_192[][BN_NIST_192_TOP] = {
        {0xFFFFFFFFFFFFFFFEULL,0xFFFFFFFFFFFFFFFDULL,0xFFFFFFFFFFFFFFFFULL},
        {0xFFFFFFFFFFFFFFFDULL,0xFFFFFFFFFFFFFFFCULL,0xFFFFFFFFFFFFFFFFULL}
        };
+static const BN_ULONG _nist_p_192_sqr[] = {
+       0x0000000000000001ULL,0x0000000000000002ULL,0x0000000000000001ULL,
+       0xFFFFFFFFFFFFFFFEULL,0xFFFFFFFFFFFFFFFDULL,0xFFFFFFFFFFFFFFFFULL
+       };
 static const BN_ULONG _nist_p_224[][BN_NIST_224_TOP] = {
        {0x0000000000000001ULL,0xFFFFFFFF00000000ULL,
         0xFFFFFFFFFFFFFFFFULL,0x00000000FFFFFFFFULL},
        {0x0000000000000002ULL,0xFFFFFFFE00000000ULL,
         0xFFFFFFFFFFFFFFFFULL,0x00000001FFFFFFFFULL} /* this one is "carry-full" */
        };
+static const BN_ULONG _nist_p_224_sqr[] = {
+       0x0000000000000001ULL,0xFFFFFFFE00000000ULL,
+       0xFFFFFFFFFFFFFFFFULL,0x0000000200000000ULL,
+       0x0000000000000000ULL,0xFFFFFFFFFFFFFFFEULL,
+       0xFFFFFFFFFFFFFFFFULL
+       };
 static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
        {0xFFFFFFFFFFFFFFFFULL,0x00000000FFFFFFFFULL,
         0x0000000000000000ULL,0xFFFFFFFF00000001ULL},
@@ -91,6 +101,12 @@ static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
        {0xFFFFFFFFFFFFFFFBULL,0x00000004FFFFFFFFULL,
         0x0000000000000000ULL,0xFFFFFFFB00000005ULL},
        };
+static const BN_ULONG _nist_p_256_sqr[] = {
+       0x0000000000000001ULL,0xFFFFFFFE00000000ULL,
+       0xFFFFFFFFFFFFFFFFULL,0x00000001FFFFFFFEULL,
+       0x00000001FFFFFFFEULL,0x00000001FFFFFFFEULL,
+       0xFFFFFFFE00000001ULL,0xFFFFFFFE00000002ULL
+       };
 static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
        {0x00000000FFFFFFFFULL,0xFFFFFFFF00000000ULL,0xFFFFFFFFFFFFFFFEULL,
         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
@@ -103,24 +119,48 @@ static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
        {0x00000004FFFFFFFBULL,0xFFFFFFFB00000000ULL,0xFFFFFFFFFFFFFFFAULL,
         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
        };
+static const BN_ULONG _nist_p_384_sqr[] = {
+       0xFFFFFFFE00000001ULL,0x0000000200000000ULL,0xFFFFFFFE00000000ULL,
+       0x0000000200000000ULL,0x0000000000000001ULL,0x0000000000000000ULL,
+       0x00000001FFFFFFFEULL,0xFFFFFFFE00000000ULL,0xFFFFFFFFFFFFFFFDULL,
+       0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL
+       };
 static const BN_ULONG _nist_p_521[] =
        {0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
        0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
        0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
        0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
        0x00000000000001FFULL};
+static const BN_ULONG _nist_p_521_sqr[] = {
+       0x0000000000000001ULL,0x0000000000000000ULL,0x0000000000000000ULL,
+       0x0000000000000000ULL,0x0000000000000000ULL,0x0000000000000000ULL,
+       0x0000000000000000ULL,0x0000000000000000ULL,0xFFFFFFFFFFFFFC00ULL,
+       0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
+       0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
+       0xFFFFFFFFFFFFFFFFULL,0x000000000003FFFFULL
+       };
 #elif BN_BITS2 == 32
 static const BN_ULONG _nist_p_192[][BN_NIST_192_TOP] = {
        {0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
        {0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
        {0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFC,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF}
        };
+static const BN_ULONG _nist_p_192_sqr[] = {
+       0x00000001,0x00000000,0x00000002,0x00000000,0x00000001,0x00000000,
+       0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF
+       };
 static const BN_ULONG _nist_p_224[][BN_NIST_224_TOP] = {
        {0x00000001,0x00000000,0x00000000,0xFFFFFFFF,
         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
        {0x00000002,0x00000000,0x00000000,0xFFFFFFFE,
         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF}
        };
+static const BN_ULONG _nist_p_224_sqr[] = {
+       0x00000001,0x00000000,0x00000000,0xFFFFFFFE,
+       0xFFFFFFFF,0xFFFFFFFF,0x00000000,0x00000002,
+       0x00000000,0x00000000,0xFFFFFFFE,0xFFFFFFFF,
+       0xFFFFFFFF,0xFFFFFFFF
+       };
 static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
        {0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0x00000000,
         0x00000000,0x00000000,0x00000001,0xFFFFFFFF},
@@ -133,6 +173,12 @@ static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
        {0xFFFFFFFB,0xFFFFFFFF,0xFFFFFFFF,0x00000004,
         0x00000000,0x00000000,0x00000005,0xFFFFFFFB},
        };
+static const BN_ULONG _nist_p_256_sqr[] = {
+       0x00000001,0x00000000,0x00000000,0xFFFFFFFE,
+       0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFE,0x00000001,
+       0xFFFFFFFE,0x00000001,0xFFFFFFFE,0x00000001,
+       0x00000001,0xFFFFFFFE,0x00000002,0xFFFFFFFE
+       };
 static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
        {0xFFFFFFFF,0x00000000,0x00000000,0xFFFFFFFF,0xFFFFFFFE,0xFFFFFFFF,
         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
@@ -145,10 +191,24 @@ static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
        {0xFFFFFFFB,0x00000004,0x00000000,0xFFFFFFFB,0xFFFFFFFA,0xFFFFFFFF,
         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
        };
+static const BN_ULONG _nist_p_384_sqr[] = {
+       0x00000001,0xFFFFFFFE,0x00000000,0x00000002,0x00000000,0xFFFFFFFE,
+       0x00000000,0x00000002,0x00000001,0x00000000,0x00000000,0x00000000,
+       0xFFFFFFFE,0x00000001,0x00000000,0xFFFFFFFE,0xFFFFFFFD,0xFFFFFFFF,
+       0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF
+       };
 static const BN_ULONG _nist_p_521[] = {0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
        0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
        0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
        0xFFFFFFFF,0x000001FF};
+static const BN_ULONG _nist_p_521_sqr[] = {
+       0x00000001,0x00000000,0x00000000,0x00000000,0x00000000,0x00000000,
+       0x00000000,0x00000000,0x00000000,0x00000000,0x00000000,0x00000000,
+       0x00000000,0x00000000,0x00000000,0x00000000,0xFFFFFC00,0xFFFFFFFF,
+       0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
+       0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
+       0xFFFFFFFF,0xFFFFFFFF,0x0003FFFF
+       };
 #else
 #error "unsupported BN_BITS2"
 #endif
@@ -231,7 +291,9 @@ static void nist_cp_bn_0(BN_ULONG *buf, BN_ULONG *a, int top, int max)
        int i;
        BN_ULONG *_tmp1 = (buf), *_tmp2 = (a);
 
+#ifdef BN_DEBUG
        OPENSSL_assert(top <= max);
+#endif
        for (i = (top); i != 0; i--)
                *_tmp1++ = *_tmp2++;
        for (i = (max) - (top); i != 0; i--)
@@ -293,11 +355,16 @@ int BN_nist_mod_192(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                 c_d[BN_NIST_192_TOP],
                *res;
        size_t   mask;
+       static const BIGNUM _bignum_nist_p_192_sqr = {
+               (BN_ULONG *)_nist_p_192_sqr,
+               sizeof(_nist_p_192_sqr)/sizeof(_nist_p_192_sqr[0]),
+               sizeof(_nist_p_192_sqr)/sizeof(_nist_p_192_sqr[0]),
+               0,BN_FLG_STATIC_DATA };
 
        field = &_bignum_nist_p_192; /* just to make sure */
 
-       if (BN_is_negative(a) || a->top > 2*BN_NIST_192_TOP)
-               return BN_nnmod(r, field, a, ctx);
+       if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_192_sqr)>=0)
+               return BN_nnmod(r, a, field, ctx);
 
        i = BN_ucmp(field, a);
        if (i == 0)
@@ -321,14 +388,14 @@ int BN_nist_mod_192(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        nist_cp_bn_0(buf, a_d + BN_NIST_192_TOP, top - BN_NIST_192_TOP, BN_NIST_192_TOP);
 
        nist_set_192(t_d, buf, 0, 3, 3);
-       carry = bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
+       carry = (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
        nist_set_192(t_d, buf, 4, 4, 0);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
        nist_set_192(t_d, buf, 5, 5, 5)
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
 
        if (carry > 0)
-               carry = bn_sub_words(r_d,r_d,_nist_p_192[carry-1],BN_NIST_192_TOP);
+               carry = (int)bn_sub_words(r_d,r_d,_nist_p_192[carry-1],BN_NIST_192_TOP);
        else
                carry = 1;
 
@@ -345,11 +412,6 @@ int BN_nist_mod_192(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        r->top = BN_NIST_192_TOP;
        bn_correct_top(r);
 
-       if (BN_ucmp(field, r) <= 0)
-               {
-               if (!BN_usub(r, r, field)) return 0;
-               }
-
        return 1;
        }
 
@@ -378,11 +440,17 @@ int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                *res;
        size_t   mask;
        union { bn_addsub_f f; size_t p; } u;
+       static const BIGNUM _bignum_nist_p_224_sqr = {
+               (BN_ULONG *)_nist_p_224_sqr,
+               sizeof(_nist_p_224_sqr)/sizeof(_nist_p_224_sqr[0]),
+               sizeof(_nist_p_224_sqr)/sizeof(_nist_p_224_sqr[0]),
+               0,BN_FLG_STATIC_DATA };
+
 
        field = &_bignum_nist_p_224; /* just to make sure */
 
-       if (BN_is_negative(a) || a->top > 2*BN_NIST_224_TOP)
-               return BN_nnmod(r, field, a, ctx);
+       if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_224_sqr)>=0)
+               return BN_nnmod(r, a, field, ctx);
 
        i = BN_ucmp(field, a);
        if (i == 0)
@@ -414,13 +482,13 @@ int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        nist_cp_bn_0(buf, a_d + BN_NIST_224_TOP, top - BN_NIST_224_TOP, BN_NIST_224_TOP);
 #endif
        nist_set_224(t_d, buf, 10, 9, 8, 7, 0, 0, 0);
-       carry = bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+       carry = (int)bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
        nist_set_224(t_d, buf, 0, 13, 12, 11, 0, 0, 0);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
        nist_set_224(t_d, buf, 13, 12, 11, 10, 9, 8, 7);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
        nist_set_224(t_d, buf, 0, 0, 0, 0, 13, 12, 11);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
 
 #if BN_BITS2==64
        carry = (int)(r_d[BN_NIST_224_TOP-1]>>32);
@@ -428,7 +496,7 @@ int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        u.f = bn_sub_words;
        if (carry > 0)
                {
-               carry = bn_sub_words(r_d,r_d,_nist_p_224[carry-1],BN_NIST_224_TOP);
+               carry = (int)bn_sub_words(r_d,r_d,_nist_p_224[carry-1],BN_NIST_224_TOP);
 #if BN_BITS2==64
                carry=(int)(~(r_d[BN_NIST_224_TOP-1]>>32))&1;
 #endif
@@ -441,7 +509,7 @@ int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                 * the modulus. but if it does, then result has
                 * to be compared to the modulus and conditionally
                 * adjusted by *subtracting* the latter. */
-               carry = bn_add_words(r_d,r_d,_nist_p_224[-carry-1],BN_NIST_224_TOP);
+               carry = (int)bn_add_words(r_d,r_d,_nist_p_224[-carry-1],BN_NIST_224_TOP);
                mask = 0-(size_t)carry;
                u.p = ((size_t)bn_sub_words&mask) | ((size_t)bn_add_words&~mask);
                }
@@ -456,11 +524,6 @@ int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        r->top = BN_NIST_224_TOP;
        bn_correct_top(r);
 
-       if (BN_ucmp(field, r) <= 0)
-               {
-               if (!BN_usub(r, r, field)) return 0;
-               }
-
        return 1;
        }
 
@@ -488,11 +551,16 @@ int BN_nist_mod_256(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                *res;
        size_t   mask;
        union { bn_addsub_f f; size_t p; } u;
+       static const BIGNUM _bignum_nist_p_256_sqr = {
+               (BN_ULONG *)_nist_p_256_sqr,
+               sizeof(_nist_p_256_sqr)/sizeof(_nist_p_256_sqr[0]),
+               sizeof(_nist_p_256_sqr)/sizeof(_nist_p_256_sqr[0]),
+               0,BN_FLG_STATIC_DATA };
 
        field = &_bignum_nist_p_256; /* just to make sure */
 
-       if (BN_is_negative(a) || a->top > 2*BN_NIST_256_TOP)
-               return BN_nnmod(r, field, a, ctx);
+       if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_256_sqr)>=0)
+               return BN_nnmod(r, a, field, ctx);
 
        i = BN_ucmp(field, a);
        if (i == 0)
@@ -519,7 +587,7 @@ int BN_nist_mod_256(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        nist_set_256(t_d, buf, 15, 14, 13, 12, 11, 0, 0, 0);
        /*S2*/
        nist_set_256(c_d, buf, 0, 15, 14, 13, 12, 0, 0, 0);
-       carry = bn_add_words(t_d, t_d, c_d, BN_NIST_256_TOP);
+       carry = (int)bn_add_words(t_d, t_d, c_d, BN_NIST_256_TOP);
        /* left shift */
                {
                register BN_ULONG *ap,t,c;
@@ -534,33 +602,33 @@ int BN_nist_mod_256(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                carry <<= 1;
                carry  |= c;
                }
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
        /*S3*/
        nist_set_256(t_d, buf, 15, 14, 0, 0, 0, 10, 9, 8);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
        /*S4*/
        nist_set_256(t_d, buf, 8, 13, 15, 14, 13, 11, 10, 9);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
        /*D1*/
        nist_set_256(t_d, buf, 10, 8, 0, 0, 0, 13, 12, 11);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
        /*D2*/
        nist_set_256(t_d, buf, 11, 9, 0, 0, 15, 14, 13, 12);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
        /*D3*/
        nist_set_256(t_d, buf, 12, 0, 10, 9, 8, 15, 14, 13);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
        /*D4*/
        nist_set_256(t_d, buf, 13, 0, 11, 10, 9, 0, 15, 14);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
 
        /* see BN_nist_mod_224 for explanation */
        u.f = bn_sub_words;
        if (carry > 0)
-               carry = bn_sub_words(r_d,r_d,_nist_p_256[carry-1],BN_NIST_256_TOP);
+               carry = (int)bn_sub_words(r_d,r_d,_nist_p_256[carry-1],BN_NIST_256_TOP);
        else if (carry < 0)
                {
-               carry = bn_add_words(r_d,r_d,_nist_p_256[-carry-1],BN_NIST_256_TOP);
+               carry = (int)bn_add_words(r_d,r_d,_nist_p_256[-carry-1],BN_NIST_256_TOP);
                mask = 0-(size_t)carry;
                u.p = ((size_t)bn_sub_words&mask) | ((size_t)bn_add_words&~mask);
                }
@@ -574,11 +642,6 @@ int BN_nist_mod_256(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        r->top = BN_NIST_256_TOP;
        bn_correct_top(r);
 
-       if (BN_ucmp(field, r) <= 0)
-               {
-               if (!BN_usub(r, r, field)) return 0;
-               }
-
        return 1;
        }
 
@@ -610,11 +673,17 @@ int BN_nist_mod_384(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                *res;
        size_t   mask;
        union { bn_addsub_f f; size_t p; } u;
+       static const BIGNUM _bignum_nist_p_384_sqr = {
+               (BN_ULONG *)_nist_p_384_sqr,
+               sizeof(_nist_p_384_sqr)/sizeof(_nist_p_384_sqr[0]),
+               sizeof(_nist_p_384_sqr)/sizeof(_nist_p_384_sqr[0]),
+               0,BN_FLG_STATIC_DATA };
+
 
        field = &_bignum_nist_p_384; /* just to make sure */
 
-       if (BN_is_negative(a) || a->top > 2*BN_NIST_384_TOP)
-               return BN_nnmod(r, field, a, ctx);
+       if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_384_sqr)>=0)
+               return BN_nnmod(r, a, field, ctx);
 
        i = BN_ucmp(field, a);
        if (i == 0)
@@ -652,39 +721,39 @@ int BN_nist_mod_384(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
                        }
                *ap=c;
                }
-       carry = bn_add_words(r_d+(128/BN_BITS2), r_d+(128/BN_BITS2), 
+       carry = (int)bn_add_words(r_d+(128/BN_BITS2), r_d+(128/BN_BITS2), 
                t_d, BN_NIST_256_TOP);
        /*S2 */
-       carry += bn_add_words(r_d, r_d, buf, BN_NIST_384_TOP);
+       carry += (int)bn_add_words(r_d, r_d, buf, BN_NIST_384_TOP);
        /*S3*/
        nist_set_384(t_d,buf,20,19,18,17,16,15,14,13,12,23,22,21);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
        /*S4*/
        nist_set_384(t_d,buf,19,18,17,16,15,14,13,12,20,0,23,0);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
        /*S5*/
        nist_set_384(t_d, buf,0,0,0,0,23,22,21,20,0,0,0,0);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
        /*S6*/
        nist_set_384(t_d,buf,0,0,0,0,0,0,23,22,21,0,0,20);
-       carry += bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
        /*D1*/
        nist_set_384(t_d,buf,22,21,20,19,18,17,16,15,14,13,12,23);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
        /*D2*/
        nist_set_384(t_d,buf,0,0,0,0,0,0,0,23,22,21,20,0);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
        /*D3*/
        nist_set_384(t_d,buf,0,0,0,0,0,0,0,23,23,0,0,0);
-       carry -= bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+       carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
 
        /* see BN_nist_mod_224 for explanation */
        u.f = bn_sub_words;
        if (carry > 0)
-               carry = bn_sub_words(r_d,r_d,_nist_p_384[carry-1],BN_NIST_384_TOP);
+               carry = (int)bn_sub_words(r_d,r_d,_nist_p_384[carry-1],BN_NIST_384_TOP);
        else if (carry < 0)
                {
-               carry = bn_add_words(r_d,r_d,_nist_p_384[-carry-1],BN_NIST_384_TOP);
+               carry = (int)bn_add_words(r_d,r_d,_nist_p_384[-carry-1],BN_NIST_384_TOP);
                mask = 0-(size_t)carry;
                u.p = ((size_t)bn_sub_words&mask) | ((size_t)bn_add_words&~mask);
                }
@@ -698,81 +767,70 @@ int BN_nist_mod_384(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        r->top = BN_NIST_384_TOP;
        bn_correct_top(r);
 
-       if (BN_ucmp(field, r) <= 0)
-               {
-               if (!BN_usub(r, r, field)) return 0;
-               }
-
        return 1;
        }
 
+#define BN_NIST_521_RSHIFT     (521%BN_BITS2)
+#define BN_NIST_521_LSHIFT     (BN_BITS2-BN_NIST_521_RSHIFT)
+#define BN_NIST_521_TOP_MASK   ((BN_ULONG)BN_MASK2>>BN_NIST_521_LSHIFT)
+
 int BN_nist_mod_521(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
        BN_CTX *ctx)
        {
-#if BN_BITS2 == 64
-#define BN_NIST_521_TOP_MASK   (BN_ULONG)0x1FF
-#elif BN_BITS2 == 32
-#define BN_NIST_521_TOP_MASK   (BN_ULONG)0x1FF
-#endif
-       int     top, ret = 0;
-       BIGNUM  *tmp;
+       int     top = a->top, i;
+       BN_ULONG *r_d, *a_d = a->d,
+                t_d[BN_NIST_521_TOP],
+                val,tmp,*res;
+       size_t  mask;
+       static const BIGNUM _bignum_nist_p_521_sqr = {
+               (BN_ULONG *)_nist_p_521_sqr,
+               sizeof(_nist_p_521_sqr)/sizeof(_nist_p_521_sqr[0]),
+               sizeof(_nist_p_521_sqr)/sizeof(_nist_p_521_sqr[0]),
+               0,BN_FLG_STATIC_DATA };
 
        field = &_bignum_nist_p_521; /* just to make sure */
 
-       if (BN_is_negative(a))
-               return BN_nnmod(r, field, a, ctx);
+       if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_521_sqr)>=0)
+               return BN_nnmod(r, a, field, ctx);
 
-       /* check whether a reduction is necessary */
-       top = a->top;
-       if (top < BN_NIST_521_TOP  || ( top == BN_NIST_521_TOP &&
-           (!(a->d[BN_NIST_521_TOP-1] & ~(BN_NIST_521_TOP_MASK)))))
+       i = BN_ucmp(field, a);
+       if (i == 0)
                {
-               int i = BN_ucmp(field, a);
-               if (i == 0)
-                       {
-                       BN_zero(r);
-                       return 1;
-                       }
-               else
-                       {
-#ifdef BN_DEBUG
-                       OPENSSL_assert(i > 0); /* because 'field' is 1111...1111 */
-#endif
-                       return (r == a)? 1 : (BN_copy(r ,a) != NULL);
-                       }
+               BN_zero(r);
+               return 1;
                }
+       else if (i > 0)
+               return (r == a)? 1 : (BN_copy(r ,a) != NULL);
 
-       if (BN_num_bits(a) > 2*521)
-               return BN_nnmod(r, field, a, ctx);
-
-       BN_CTX_start(ctx);
-       tmp = BN_CTX_get(ctx);
-       if (!tmp)
-               goto err;
-
-       if (!bn_wexpand(tmp, BN_NIST_521_TOP))
-               goto err;
-       nist_cp_bn(tmp->d, a->d, BN_NIST_521_TOP);
-
-       tmp->top = BN_NIST_521_TOP;
-        tmp->d[BN_NIST_521_TOP-1]  &= BN_NIST_521_TOP_MASK;
-       bn_correct_top(tmp);
-
-       if (!BN_rshift(r, a, 521))
-               goto err;
-
-       if (!BN_uadd(r, tmp, r))
-               goto err;
-
-       if (BN_ucmp(field, r) <= 0)
+       if (r != a)
                {
-               if (!BN_usub(r, r, field)) goto err;
+               if (!bn_wexpand(r,BN_NIST_521_TOP))
+                       return 0;
+               r_d = r->d;
+               nist_cp_bn(r_d,a_d, BN_NIST_521_TOP);
                }
+       else
+               r_d = a_d;
 
-       ret = 1;
-err:
-       BN_CTX_end(ctx);
+       /* upper 521 bits, copy ... */
+       nist_cp_bn_0(t_d,a_d + (BN_NIST_521_TOP-1), top - (BN_NIST_521_TOP-1),BN_NIST_521_TOP);
+       /* ... and right shift */
+       for (val=t_d[0],i=0; i<BN_NIST_521_TOP-1; i++)
+               {
+               tmp = val>>BN_NIST_521_RSHIFT;
+               val = t_d[i+1];
+               t_d[i] = (tmp | val<<BN_NIST_521_LSHIFT) & BN_MASK2;
+               }
+       t_d[i] = val>>BN_NIST_521_RSHIFT;
+       /* lower 521 bits */
+       r_d[i] &= BN_NIST_521_TOP_MASK;
+
+       bn_add_words(r_d,r_d,t_d,BN_NIST_521_TOP);
+       mask = 0-(size_t)bn_sub_words(t_d,r_d,_nist_p_521,BN_NIST_521_TOP);
+       res  = (BN_ULONG *)(((size_t)t_d&~mask) | ((size_t)r_d&mask));
+       nist_cp_bn(r_d,res,BN_NIST_521_TOP);
+       r->top = BN_NIST_521_TOP;
+       bn_correct_top(r);
 
-       bn_check_top(r);
-       return ret;
+       return 1;
        }