Xuyi Wang / wolfSSL

Dependents:   OS

Revision:
14:167253f4e170
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/wolfcrypt/src/sp_int.c	Sat Aug 18 22:19:52 2018 +0000
@@ -0,0 +1,665 @@
+/* sp_int.c
+ *
+ * Copyright (C) 2006-2017 wolfSSL Inc.
+ *
+ * This file is part of wolfSSL.
+ *
+ * wolfSSL is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * wolfSSL is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1335, USA
+ */
+
+/* Implementation by Sean Parkinson. */
+
+#ifdef HAVE_CONFIG_H
+    #include <config.h>
+#endif
+
+#include <wolfssl/wolfcrypt/settings.h>
+#include <wolfssl/wolfcrypt/error-crypt.h>
+#ifdef NO_INLINE
+    #include <wolfssl/wolfcrypt/misc.h>
+#else
+    #define WOLFSSL_MISC_INCLUDED
+    #include <wolfcrypt/src/misc.c>
+#endif
+
+
+#ifdef WOLFSSL_SP_MATH
+
+#include <wolfssl/wolfcrypt/sp_int.h>
+
+/* Initialize the big number to be zero.
+ *
+ * a  SP integer.
+ * returns MP_OKAY always.
+ */
+int sp_init(sp_int* a)
+{
+    a->used = 0;
+    a->size = SP_INT_DIGITS;
+
+    return MP_OKAY;
+}
+
+/* Initialize up to six big numbers to be zero.
+ *
+ * a  SP integer.
+ * b  SP integer.
+ * c  SP integer.
+ * d  SP integer.
+ * e  SP integer.
+ * f  SP integer.
+ * returns MP_OKAY always.
+ */
+int sp_init_multi(sp_int* a, sp_int* b, sp_int* c, sp_int* d, sp_int* e,
+                  sp_int* f)
+{
+    if (a != NULL) {
+        a->used = 0;
+        a->size = SP_INT_DIGITS;
+    }
+    if (b != NULL) {
+        b->used = 0;
+        b->size = SP_INT_DIGITS;
+    }
+    if (c != NULL) {
+        c->used = 0;
+        c->size = SP_INT_DIGITS;
+    }
+    if (d != NULL) {
+        d->used = 0;
+        d->size = SP_INT_DIGITS;
+    }
+    if (e != NULL) {
+        e->used = 0;
+        e->size = SP_INT_DIGITS;
+    }
+    if (f != NULL) {
+        f->used = 0;
+        f->size = SP_INT_DIGITS;
+    }
+
+    return MP_OKAY;
+}
+
+/* Clear the data from the big number and set to zero.
+ *
+ * a  SP integer.
+ */
+void sp_clear(sp_int* a)
+{
+    int i;
+
+    for (i=0; i<a->used; i++)
+        a->dp[i] = 0;
+    a->used = 0;
+}
+
+/* Calculate the number of 8-bit values required to represent the big number.
+ *
+ * a  SP integer.
+ * returns the count.
+ */
+int sp_unsigned_bin_size(sp_int* a)
+{
+    int size = sp_count_bits(a);
+    return (size + 7) / 8;
+}
+
+/* Convert a number as an array of bytes in big-endian format to a big number.
+ *
+ * a     SP integer.
+ * in    Array of bytes.
+ * inSz  Number of data bytes in array.
+ * returns MP_OKAY always.
+ */
+int sp_read_unsigned_bin(sp_int* a, const byte* in, word32 inSz)
+{
+    int i, j = 0, s = 0;
+
+    a->dp[0] = 0;
+    for (i = inSz-1; i >= 0; i--) {
+        a->dp[j] |= ((sp_int_digit)in[i]) << s;
+        if (s == DIGIT_BIT - 8) {
+            a->dp[++j] = 0;
+            s = 0;
+        }
+        else if (s > DIGIT_BIT - 8) {
+            s = DIGIT_BIT - s;
+            if (j + 1 >= a->size)
+                break;
+            a->dp[++j] = in[i] >> s;
+            s = 8 - s;
+        }
+        else
+            s += 8;
+    }
+
+    a->used = j + 1;
+    if (a->dp[j] == 0)
+        a->used--;
+
+    for (j++; j < a->size; j++)
+        a->dp[j] = 0;
+
+    return MP_OKAY;
+}
+
+/* Convert a number as string in big-endian format to a big number.
+ * Only supports base-16 (hexadecimal).
+ * Negative values not supported.
+ *
+ * a      SP integer.
+ * in     NUL terminated string.
+ * radix  Number of values in a digit.
+ * returns BAD_FUNC_ARG when radix not supported or value is negative, MP_VAL
+ * when a character is not valid and MP_OKAY otherwise.
+ */
+int sp_read_radix(sp_int* a, const char* in, int radix)
+{
+    int     i, j, k;
+    char    ch;
+
+    if (radix != 16)
+        return BAD_FUNC_ARG;
+
+    if (*in == '-') {
+        return BAD_FUNC_ARG;
+    }
+
+    j = 0;
+    k = 0;
+    a->dp[0] = 0;
+    for (i = (int)(XSTRLEN(in) - 1); i >= 0; i--) {
+        ch = in[i];
+        if (ch >= '0' && ch <= '9')
+            ch -= '0';
+        else if (ch >= 'A' && ch <= 'F')
+            ch -= 'A' - 10;
+        else if (ch >= 'a' && ch <= 'f')
+            ch -= 'a' - 10;
+        else
+            return MP_VAL;
+
+        a->dp[k] |= ((sp_int_digit)ch) << j;
+        j += 4;
+        if (j == DIGIT_BIT && k < SP_INT_DIGITS)
+            a->dp[++k] = 0;
+        j &= DIGIT_BIT - 1;
+    }
+
+    a->used = k + 1;
+    if (a->dp[k] == 0)
+        a->used--;
+
+    for (k++; k < a->size; k++)
+        a->dp[k] = 0;
+
+    return MP_OKAY;
+}
+
+/* Compare two big numbers.
+ *
+ * a  SP integer.
+ * b  SP integer.
+ * returns MP_GT if a is greater than b, MP_LT if a is less than b and MP_EQ
+ * when a equals b.
+ */
+int sp_cmp(sp_int* a, sp_int* b)
+{
+    int i;
+
+    if (a->used > b->used)
+        return MP_GT;
+    else if (a->used < b->used)
+        return MP_LT;
+
+    for (i = a->used - 1; i >= 0; i--) {
+        if (a->dp[i] > b->dp[i])
+            return MP_GT;
+        else if (a->dp[i] < b->dp[i])
+            return MP_LT;
+    }
+    return MP_EQ;
+}
+
+/* Count the number of bits in the big number.
+ *
+ * a  SP integer.
+ * returns the number of bits.
+ */
+int sp_count_bits(sp_int* a)
+{
+    int r = 0;
+    sp_int_digit d;
+
+    r = a->used - 1;
+    while (r >= 0 && a->dp[r] == 0)
+        r--;
+    if (r < 0)
+        r = 0;
+    else {
+        d = a->dp[r];
+        r *= DIGIT_BIT;
+        while (d != 0) {
+            r++;
+            d >>= 1;
+        }
+    }
+
+    return r;
+}
+
+/* Determine if the most significant byte of the encoded big number as the top
+ * bit set.
+ *
+ * a  SP integer.
+ * returns 1 when the top bit is set and 0 otherwise.
+ */
+int sp_leading_bit(sp_int* a)
+{
+    int bit = 0;
+    sp_int_digit d;
+
+    if (a->used > 0) {
+        d = a->dp[a->used - 1];
+        while (d > (sp_int_digit)0xff)
+            d >>= 8;
+        bit = (int)(d >> 7);
+    }
+
+    return bit;
+}
+
+/* Convert the big number to an array of bytes in big-endian format.
+ * The array must be large enough for encoded number - use mp_unsigned_bin_size
+ * to calculate the number of bytes required.
+ *
+ * a  SP integer.
+ * returns MP_OKAY always.
+ */
+int sp_to_unsigned_bin(sp_int* a, byte* out)
+{
+    int i, j, b;
+
+    j = sp_unsigned_bin_size(a) - 1;
+    for (i=0; j>=0; i++) {
+        for (b = 0; b < SP_WORD_SIZE; b += 8) {
+            out[j--] = a->dp[i] >> b;
+            if (j < 0)
+                break;
+        }
+    }
+
+    return MP_OKAY;
+}
+
+/* Ensure the data in the big number is zeroed.
+ *
+ * a  SP integer.
+ */
+void sp_forcezero(sp_int* a)
+{
+    ForceZero(a->dp, a->used * sizeof(sp_int_digit));
+    a->used = 0;
+}
+
+/* Copy value of big number a into b.
+ *
+ * a  SP integer.
+ * b  SP integer.
+ * returns MP_OKAY always.
+ */
+int sp_copy(sp_int* a, sp_int* b)
+{
+    if (a != b) {
+        XMEMCPY(b->dp, a->dp, a->used * sizeof(sp_int_digit));
+        b->used = a->used;
+    }
+    return MP_OKAY;
+}
+
+/* Set the big number to be the value of the digit.
+ *
+ * a  SP integer.
+ * d  Digit to be set.
+ * returns MP_OKAY always.
+ */
+int sp_set(sp_int* a, sp_int_digit d)
+{
+    a->dp[0] = d;
+    a->used = 1;
+    return MP_OKAY;
+}
+
+/* Checks whether the value of the big number is zero.
+ *
+ * a  SP integer.
+ * returns 1 when value is zero and 0 otherwise.
+ */
+int sp_iszero(sp_int* a)
+{
+    return a->used == 0;
+}
+
+/* Recalculate the number of digits used.
+ *
+ * a  SP integer.
+ */
+void sp_clamp(sp_int* a)
+{
+    int i;
+
+    for (i = a->used - 1; i >= 0 && a->dp[i] == 0; i--) {
+    }
+    a->used = i + 1;
+}
+
+/* Grow big number to be able to hold l digits.
+ * This function does nothing as the number of digits is fixed.
+ *
+ * a  SP integer.
+ * l  Number of digits.
+ * retuns MP_MEM if the number of digits requested is more than available and
+ * MP_OKAY otherwise.
+ */
+int sp_grow(sp_int* a, int l)
+{
+    if (l > a->size)
+        return MP_MEM;
+    (void)a;
+    (void)l;
+    return MP_OKAY;
+}
+
+/* Sub a one digit number from the big number.
+ *
+ * a  SP integer.
+ * d  Digit to subtract.
+ * r  SP integer - result.
+ * returns MP_OKAY always.
+ */
+int sp_sub_d(sp_int* a, sp_int_digit d, sp_int* r)
+{
+    int i = 0;
+
+    r->used = a->used;
+    r->dp[0] = a->dp[0] - d;
+    if (r->dp[i] > a->dp[i]) {
+        for (; i < a->used; i++) {
+            r->dp[i] = a->dp[i] - 1;
+            if (r->dp[i] != (sp_int_digit)-1)
+               break;
+        }
+    }
+    for (; i < a->used; i++)
+        r->dp[i] = a->dp[i];
+
+    return MP_OKAY;
+}
+
+/* Compare a one digit number with a big number.
+ *
+ * a  SP integer.
+ * d  Digit to compare with.
+ * returns MP_GT if a is greater than d, MP_LT if a is less than d and MP_EQ
+ * when a equals d.
+ */
+int sp_cmp_d(sp_int *a, sp_int_digit d)
+{
+    /* special case for zero*/
+    if (a->used == 0) {
+        if (d == 0)
+            return MP_EQ;
+        else
+            return MP_LT;
+    }
+    else if (a->used > 1)
+        return MP_GT;
+
+    /* compare the only digit of a to d */
+    if (a->dp[0] > d)
+        return MP_GT;
+    else if (a->dp[0] < d)
+        return MP_LT;
+    return MP_EQ;
+}
+
+/* Left shift the number by number of bits.
+ * Bits may be larger than the word size.
+ *
+ * a  SP integer.
+ * n  Number of bits to shift.
+ * returns MP_OKAY always.
+ */
+static int sp_lshb(sp_int* a, int n)
+{
+    int i;
+
+    if (n >= SP_WORD_SIZE) {
+        sp_lshd(a, n / SP_WORD_SIZE);
+        n %= SP_WORD_SIZE;
+    }
+
+    if (n == 0)
+        return MP_OKAY;
+
+    a->dp[a->used] = 0;
+    for (i = a->used - 1; i >= 0; i--) {
+        a->dp[i+1] |= a->dp[i] >> (SP_WORD_SIZE - n);
+        a->dp[i] = a->dp[i] << n;
+    }
+    if (a->dp[a->used] != 0)
+        a->used++;
+
+    return MP_OKAY;
+}
+
+/* Subtract two large numbers into result: r = a - b
+ * a must be greater than b.
+ *
+ * a  SP integer.
+ * b  SP integer.
+ * r  SP integer.
+ * returns MP_OKAY always.
+ */
+static int sp_sub(sp_int* a, sp_int* b, sp_int* r)
+{
+    int i;
+    sp_int_digit c = 0;
+    sp_int_digit t;
+
+    for (i = 0; i < a->used && i < b->used; i++) {
+        t = a->dp[i] - b->dp[i] - c;
+        if (c == 0)
+            c = t > a->dp[i];
+        else
+            c = t >= a->dp[i];
+        r->dp[i] = t;
+    }
+    for (; i < a->used; i++) {
+        r->dp[i] = a->dp[i] - c;
+        c = r->dp[i] == (sp_int_digit)-1;
+    }
+    r->used = i;
+    sp_clamp(r);
+
+    return MP_OKAY;
+}
+
+/* Calculate the r = a mod m.
+ *
+ * a  SP integer.
+ * m  SP integer.
+ * r  SP integer.
+ * returns MP_OKAY always.
+ */
+int sp_mod(sp_int* a, sp_int* m, sp_int* r)
+{
+    sp_int t;
+    int mBits = sp_count_bits(m);
+    int rBits;
+
+    if (a != r)
+        sp_copy(a, r);
+    sp_init(&t);
+
+    rBits = sp_count_bits(r);
+    while (rBits > mBits) {
+        sp_copy(m, &t);
+        sp_lshb(&t, rBits - mBits);
+
+        if (sp_cmp(&t, r) == MP_GT) {
+            sp_copy(m, &t);
+            sp_lshb(&t, rBits - mBits - 1);
+        }
+        sp_sub(r, &t, r);
+
+        rBits = sp_count_bits(r);
+    }
+    if (sp_cmp(r, m) != MP_LT)
+        sp_sub(r, m, r);
+
+    return MP_OKAY;
+}
+
+#if defined(USE_FAST_MATH) || !defined(NO_BIG_INT)
+/* Clear all data in the big number and sets value to zero.
+ *
+ * a  SP integer.
+ */
+void sp_zero(sp_int* a)
+{
+    XMEMSET(a->dp, 0, a->size);
+    a->used = 0;
+}
+
+/* Add a one digit number to the big number.
+ *
+ * a  SP integer.
+ * d  Digit to add.
+ * r  SP integer - result.
+ * returns MP_OKAY always.
+ */
+int sp_add_d(sp_int* a, sp_int_digit d, sp_int* r)
+{
+    int i = 0;
+
+    r->used = a->used;
+    r->dp[0] = a->dp[0] + d;
+    if (r->dp[i] < a->dp[i]) {
+        for (; i < a->used; i++) {
+            r->dp[i] = a->dp[i] + 1;
+            if (r->dp[i] != 0)
+               break;
+        }
+
+        if (i == a->used) {
+            r->used++;
+            r->dp[i] = 1;
+        }
+    }
+    for (; i < a->used; i++)
+        r->dp[i] = a->dp[i];
+
+    return MP_OKAY;
+}
+
+/* Left shift the big number by a number of digits.
+ * WIll chop off digits overflowing maximum size.
+ *
+ * a  SP integer.
+ * s  Number of digits to shift.
+ * returns MP_OKAY always.
+ */
+int sp_lshd(sp_int* a, int s)
+{
+    if (a->used + s > a->size)
+        a->used = a->size - s;
+
+    XMEMMOVE(a->dp + s, a->dp, a->used * SP_INT_DIGITS);
+    a->used += s;
+    XMEMSET(a->dp, 0, s * sizeof(sp_int_digit));
+
+    return MP_OKAY;
+}
+#endif
+
+#ifndef NO_PWDBASED
+/* Add two large numbers into result: r = a + b
+ *
+ * a  SP integer.
+ * b  SP integer.
+ * r  SP integer.
+ * returns MP_OKAY always.
+ */
+int sp_add(sp_int* a, sp_int* b, sp_int* r)
+{
+    int i;
+    sp_digit c = 0;
+    sp_digit t;
+
+    for (i = 0; i < a->used && i < b->used; i++) {
+        t = a->dp[i] + b->dp[i] + c;
+        if (c == 0)
+            c = t < a->dp[i];
+        else
+            c = t <= a->dp[i];
+        r->dp[i] = t;
+    }
+    for (; i < a->used; i++) {
+        r->dp[i] = a->dp[i] + c;
+        c = r->dp[i] == 0;
+    }
+    for (; i < b->used; i++) {
+        r->dp[i] = b->dp[i] + c;
+        c = r->dp[i] == 0;
+    }
+    r->dp[i] = c;
+    r->used = (int)(i + c);
+
+    return MP_OKAY;
+}
+#endif
+
+#ifndef NO_RSA
+/* Set a number into the big number.
+ *
+ * a  SP integer.
+ * b  Value to set.
+ * returns MP_OKAY always.
+ */
+int sp_set_int(sp_int* a, unsigned long b)
+{
+    a->used = 1;
+    a->dp[0] = b;
+
+    return MP_OKAY;
+}
+#endif
+
+#if !defined(USE_FAST_MATH)
+/* Returns the run time settings.
+ *
+ * returns the settings value.
+ */
+word32 CheckRunTimeSettings(void)
+{
+    return CTC_SETTINGS;
+}
+#endif
+
+#endif
+
+