/* * BigInt * (C) 1999-2008,2012 Jack Lloyd * 2007 FlexSecure * * Botan is released under the Simplified BSD License (see license.txt) */ #ifndef BOTAN_BIGINT_H__ #define BOTAN_BIGINT_H__ #include #include #include #include #include namespace Botan { /** * Arbitrary precision integer */ class BOTAN_DLL BigInt { public: /** * Base enumerator for encoding and decoding */ enum Base { Decimal = 10, Hexadecimal = 16, Binary = 256 }; /** * Sign symbol definitions for positive and negative numbers */ enum Sign { Negative = 0, Positive = 1 }; /** * DivideByZero Exception */ struct BOTAN_DLL DivideByZero : public Exception { DivideByZero() : Exception("BigInt divide by zero") {} }; /** * Create empty BigInt */ BigInt() {} /** * Create BigInt from 64 bit integer * @param n initial value of this BigInt */ BigInt(u64bit n); /** * Copy Constructor * @param other the BigInt to copy */ BigInt(const BigInt& other); /** * Create BigInt from a string. If the string starts with 0x the * rest of the string will be interpreted as hexadecimal digits. * Otherwise, it will be interpreted as a decimal number. * * @param str the string to parse for an integer value */ BigInt(const std::string& str); /** * Create a BigInt from an integer in a byte array * @param buf the byte array holding the value * @param length size of buf * @param base is the number base of the integer in buf */ BigInt(const byte buf[], size_t length, Base base = Binary); /** * \brief Create a random BigInt of the specified size * * @param rng random number generator * @param bits size in bits * @param set_high_bit if true, the highest bit is always set * * @see randomize */ BigInt(RandomNumberGenerator& rng, size_t bits, bool set_high_bit = true); /** * Create BigInt of specified size, all zeros * @param sign the sign * @param n size of the internal register in words */ BigInt(Sign sign, size_t n); /** * Move constructor */ BigInt(BigInt&& other) { this->swap(other); } /** * Move assignment */ BigInt& operator=(BigInt&& other) { if(this != &other) this->swap(other); return (*this); } /** * Copy assignment */ BigInt& operator=(const BigInt&) = default; /** * Swap this value with another * @param other BigInt to swap values with */ void swap(BigInt& other) { m_reg.swap(other.m_reg); std::swap(m_signedness, other.m_signedness); } void swap_reg(secure_vector& reg) { m_reg.swap(reg); } /** * += operator * @param y the BigInt to add to this */ BigInt& operator+=(const BigInt& y); /** * -= operator * @param y the BigInt to subtract from this */ BigInt& operator-=(const BigInt& y); /** * *= operator * @param y the BigInt to multiply with this */ BigInt& operator*=(const BigInt& y); /** * /= operator * @param y the BigInt to divide this by */ BigInt& operator/=(const BigInt& y); /** * Modulo operator * @param y the modulus to reduce this by */ BigInt& operator%=(const BigInt& y); /** * Modulo operator * @param y the modulus (word) to reduce this by */ word operator%=(word y); /** * Left shift operator * @param shift the number of bits to shift this left by */ BigInt& operator<<=(size_t shift); /** * Right shift operator * @param shift the number of bits to shift this right by */ BigInt& operator>>=(size_t shift); /** * Increment operator */ BigInt& operator++() { return (*this += 1); } /** * Decrement operator */ BigInt& operator--() { return (*this -= 1); } /** * Postfix increment operator */ BigInt operator++(int) { BigInt x = (*this); ++(*this); return x; } /** * Postfix decrement operator */ BigInt operator--(int) { BigInt x = (*this); --(*this); return x; } /** * Unary negation operator * @return negative this */ BigInt operator-() const; /** * ! operator * @return true iff this is zero, otherwise false */ bool operator !() const { return (!is_nonzero()); } /** * Zeroize the BigInt. The size of the underlying register is not * modified. */ void clear() { zeroise(m_reg); } /** * Compare this to another BigInt * @param n the BigInt value to compare with * @param check_signs include sign in comparison? * @result if (thisn) return 1, if both * values are identical return 0 [like Perl's <=> operator] */ s32bit cmp(const BigInt& n, bool check_signs = true) const; /** * Test if the integer has an even value * @result true if the integer is even, false otherwise */ bool is_even() const { return (get_bit(0) == 0); } /** * Test if the integer has an odd value * @result true if the integer is odd, false otherwise */ bool is_odd() const { return (get_bit(0) == 1); } /** * Test if the integer is not zero * @result true if the integer is non-zero, false otherwise */ bool is_nonzero() const { return (!is_zero()); } /** * Test if the integer is zero * @result true if the integer is zero, false otherwise */ bool is_zero() const { const size_t sw = sig_words(); for(size_t i = 0; i != sw; ++i) if(m_reg[i]) return false; return true; } /** * Set bit at specified position * @param n bit position to set */ void set_bit(size_t n); /** * Clear bit at specified position * @param n bit position to clear */ void clear_bit(size_t n); /** * Clear all but the lowest n bits * @param n amount of bits to keep */ void mask_bits(size_t n) { if(n == 0) { clear(); return; } const size_t top_word = n / BOTAN_MP_WORD_BITS; const word mask = (static_cast(1) << (n % BOTAN_MP_WORD_BITS)) - 1; if(top_word < size()) { const size_t len = size() - (top_word + 1); if (len > 0) { clear_mem(&m_reg[top_word+1], len); } m_reg[top_word] &= mask; } } /** * Return bit value at specified position * @param n the bit offset to test * @result true, if the bit at position n is set, false otherwise */ bool get_bit(size_t n) const { return ((word_at(n / BOTAN_MP_WORD_BITS) >> (n % BOTAN_MP_WORD_BITS)) & 1); } /** * Return (a maximum of) 32 bits of the complete value * @param offset the offset to start extracting * @param length amount of bits to extract (starting at offset) * @result the integer extracted from the register starting at * offset with specified length */ u32bit get_substring(size_t offset, size_t length) const; /** * Convert this value into a u32bit, if it is in the range * [0 ... 2**32-1], or otherwise throw an exception. * @result the value as a u32bit if conversion is possible */ u32bit to_u32bit() const; /** * @param n the offset to get a byte from * @result byte at offset n */ byte byte_at(size_t n) const { return get_byte(sizeof(word) - (n % sizeof(word)) - 1, word_at(n / sizeof(word))); } /** * Return the word at a specified position of the internal register * @param n position in the register * @return value at position n */ word word_at(size_t n) const { return ((n < size()) ? m_reg[n] : 0); } void set_word_at(size_t i, word w) { grow_to(i + 1); m_reg[i] = w; } /** * Tests if the sign of the integer is negative * @result true, iff the integer has a negative sign */ bool is_negative() const { return (sign() == Negative); } /** * Tests if the sign of the integer is positive * @result true, iff the integer has a positive sign */ bool is_positive() const { return (sign() == Positive); } /** * Return the sign of the integer * @result the sign of the integer */ Sign sign() const { return (m_signedness); } /** * @result the opposite sign of the represented integer value */ Sign reverse_sign() const; /** * Flip the sign of this BigInt */ void flip_sign(); /** * Set sign of the integer * @param sign new Sign to set */ void set_sign(Sign sign); /** * @result absolute (positive) value of this */ BigInt abs() const; /** * Give size of internal register * @result size of internal register in words */ size_t size() const { return m_reg.size(); } /** * Return how many words we need to hold this value * @result significant words of the represented integer value */ size_t sig_words() const { const word* x = m_reg.data(); size_t sig = m_reg.size(); while(sig && (x[sig-1] == 0)) sig--; return sig; } /** * Give byte length of the integer * @result byte length of the represented integer value */ size_t bytes() const; /** * Get the bit length of the integer * @result bit length of the represented integer value */ size_t bits() const; /** * Return a mutable pointer to the register * @result a pointer to the start of the internal register */ word* mutable_data() { return m_reg.data(); } /** * Return a const pointer to the register * @result a pointer to the start of the internal register */ const word* data() const { return m_reg.data(); } secure_vector& get_word_vector() { return m_reg; } const secure_vector& get_word_vector() const { return m_reg; } /** * Increase internal register buffer to at least n words * @param n new size of register */ void grow_to(size_t n); /** * Fill BigInt with a random number with size of bitsize * * If \p set_high_bit is true, the highest bit will be set, which causes * the entropy to be \a bits-1. Otherwise the highest bit is randomly chosen * by the rng, causing the entropy to be \a bits. * * @param rng the random number generator to use * @param bitsize number of bits the created random value should have * @param set_high_bit if true, the highest bit is always set */ void randomize(RandomNumberGenerator& rng, size_t bitsize, bool set_high_bit = true); /** * Store BigInt-value in a given byte array * @param buf destination byte array for the integer value */ void binary_encode(byte buf[]) const; /** * Read integer value from a byte array with given size * @param buf byte array buffer containing the integer * @param length size of buf */ void binary_decode(const byte buf[], size_t length); /** * Read integer value from a byte array (secure_vector) * @param buf the array to load from */ void binary_decode(const secure_vector& buf) { binary_decode(buf.data(), buf.size()); } /** * @param base the base to measure the size for * @return size of this integer in base base */ size_t encoded_size(Base base = Binary) const; /** * @param rng a random number generator * @param min the minimum value * @param max the maximum value * @return random integer in [min,max) */ static BigInt random_integer(RandomNumberGenerator& rng, const BigInt& min, const BigInt& max); /** * Create a power of two * @param n the power of two to create * @return bigint representing 2^n */ static BigInt power_of_2(size_t n) { BigInt b; b.set_bit(n); return b; } /** * Encode the integer value from a BigInt to a std::vector of bytes * @param n the BigInt to use as integer source * @param base number-base of resulting byte array representation * @result secure_vector of bytes containing the integer with given base */ static std::vector encode(const BigInt& n, Base base = Binary); /** * Encode the integer value from a BigInt to a secure_vector of bytes * @param n the BigInt to use as integer source * @param base number-base of resulting byte array representation * @result secure_vector of bytes containing the integer with given base */ static secure_vector encode_locked(const BigInt& n, Base base = Binary); /** * Encode the integer value from a BigInt to a byte array * @param buf destination byte array for the encoded integer * value with given base * @param n the BigInt to use as integer source * @param base number-base of resulting byte array representation */ static void encode(byte buf[], const BigInt& n, Base base = Binary); /** * Create a BigInt from an integer in a byte array * @param buf the binary value to load * @param length size of buf * @param base number-base of the integer in buf * @result BigInt representing the integer in the byte array */ static BigInt decode(const byte buf[], size_t length, Base base = Binary); /** * Create a BigInt from an integer in a byte array * @param buf the binary value to load * @param base number-base of the integer in buf * @result BigInt representing the integer in the byte array */ static BigInt decode(const secure_vector& buf, Base base = Binary) { return BigInt::decode(buf.data(), buf.size(), base); } /** * Create a BigInt from an integer in a byte array * @param buf the binary value to load * @param base number-base of the integer in buf * @result BigInt representing the integer in the byte array */ static BigInt decode(const std::vector& buf, Base base = Binary) { return BigInt::decode(buf.data(), buf.size(), base); } /** * Encode a BigInt to a byte array according to IEEE 1363 * @param n the BigInt to encode * @param bytes the length of the resulting secure_vector * @result a secure_vector containing the encoded BigInt */ static secure_vector encode_1363(const BigInt& n, size_t bytes); static void encode_1363(byte out[], size_t bytes, const BigInt& n); private: secure_vector m_reg; Sign m_signedness = Positive; }; /* * Arithmetic Operators */ BigInt BOTAN_DLL operator+(const BigInt& x, const BigInt& y); BigInt BOTAN_DLL operator-(const BigInt& x, const BigInt& y); BigInt BOTAN_DLL operator*(const BigInt& x, const BigInt& y); BigInt BOTAN_DLL operator/(const BigInt& x, const BigInt& d); BigInt BOTAN_DLL operator%(const BigInt& x, const BigInt& m); word BOTAN_DLL operator%(const BigInt& x, word m); BigInt BOTAN_DLL operator<<(const BigInt& x, size_t n); BigInt BOTAN_DLL operator>>(const BigInt& x, size_t n); /* * Comparison Operators */ inline bool operator==(const BigInt& a, const BigInt& b) { return (a.cmp(b) == 0); } inline bool operator!=(const BigInt& a, const BigInt& b) { return (a.cmp(b) != 0); } inline bool operator<=(const BigInt& a, const BigInt& b) { return (a.cmp(b) <= 0); } inline bool operator>=(const BigInt& a, const BigInt& b) { return (a.cmp(b) >= 0); } inline bool operator<(const BigInt& a, const BigInt& b) { return (a.cmp(b) < 0); } inline bool operator>(const BigInt& a, const BigInt& b) { return (a.cmp(b) > 0); } /* * I/O Operators */ BOTAN_DLL std::ostream& operator<<(std::ostream&, const BigInt&); BOTAN_DLL std::istream& operator>>(std::istream&, BigInt&); } namespace std { template<> inline void swap(Botan::BigInt& x, Botan::BigInt& y) { x.swap(y); } } #endif