Source file src/crypto/internal/fips140/bigmod/nat.go

     1  // Copyright 2021 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package bigmod
     6  
     7  import (
     8  	_ "crypto/internal/fips140/check"
     9  	"crypto/internal/fips140deps/byteorder"
    10  	"errors"
    11  	"math/bits"
    12  )
    13  
    14  const (
    15  	// _W is the size in bits of our limbs.
    16  	_W = bits.UintSize
    17  	// _S is the size in bytes of our limbs.
    18  	_S = _W / 8
    19  )
    20  
    21  // Note: These functions make many loops over all the words in a Nat.
    22  // These loops used to be in assembly, invisible to -race, -asan, and -msan,
    23  // but now they are in Go and incur significant overhead in those modes.
    24  // To bring the old performance back, we mark all functions that loop
    25  // over Nat words with //go:norace. Because //go:norace does not
    26  // propagate across inlining, we must also mark functions that inline
    27  // //go:norace functions - specifically, those that inline add, addMulVVW,
    28  // assign, cmpGeq, rshift1, and sub.
    29  
    30  // choice represents a constant-time boolean. The value of choice is always
    31  // either 1 or 0. We use an int instead of bool in order to make decisions in
    32  // constant time by turning it into a mask.
    33  type choice uint
    34  
    35  func not(c choice) choice { return 1 ^ c }
    36  
    37  const yes = choice(1)
    38  const no = choice(0)
    39  
    40  // ctMask is all 1s if on is yes, and all 0s otherwise.
    41  func ctMask(on choice) uint { return -uint(on) }
    42  
    43  // ctEq returns 1 if x == y, and 0 otherwise. The execution time of this
    44  // function does not depend on its inputs.
    45  func ctEq(x, y uint) choice {
    46  	// If x != y, then either x - y or y - x will generate a carry.
    47  	_, c1 := bits.Sub(x, y, 0)
    48  	_, c2 := bits.Sub(y, x, 0)
    49  	return not(choice(c1 | c2))
    50  }
    51  
    52  // Nat represents an arbitrary natural number
    53  //
    54  // Each Nat has an announced length, which is the number of limbs it has stored.
    55  // Operations on this number are allowed to leak this length, but will not leak
    56  // any information about the values contained in those limbs.
    57  type Nat struct {
    58  	// limbs is little-endian in base 2^W with W = bits.UintSize.
    59  	limbs []uint
    60  }
    61  
    62  // preallocTarget is the size in bits of the numbers used to implement the most
    63  // common and most performant RSA key size. It's also enough to cover some of
    64  // the operations of key sizes up to 4096.
    65  const preallocTarget = 2048
    66  const preallocLimbs = (preallocTarget + _W - 1) / _W
    67  
    68  // NewNat returns a new nat with a size of zero, just like new(Nat), but with
    69  // the preallocated capacity to hold a number of up to preallocTarget bits.
    70  // NewNat inlines, so the allocation can live on the stack.
    71  func NewNat() *Nat {
    72  	limbs := make([]uint, 0, preallocLimbs)
    73  	return &Nat{limbs}
    74  }
    75  
    76  // expand expands x to n limbs, leaving its value unchanged.
    77  func (x *Nat) expand(n int) *Nat {
    78  	if len(x.limbs) > n {
    79  		panic("bigmod: internal error: shrinking nat")
    80  	}
    81  	if cap(x.limbs) < n {
    82  		newLimbs := make([]uint, n)
    83  		copy(newLimbs, x.limbs)
    84  		x.limbs = newLimbs
    85  		return x
    86  	}
    87  	extraLimbs := x.limbs[len(x.limbs):n]
    88  	clear(extraLimbs)
    89  	x.limbs = x.limbs[:n]
    90  	return x
    91  }
    92  
    93  // reset returns a zero nat of n limbs, reusing x's storage if n <= cap(x.limbs).
    94  func (x *Nat) reset(n int) *Nat {
    95  	if cap(x.limbs) < n {
    96  		x.limbs = make([]uint, n)
    97  		return x
    98  	}
    99  	// Clear both the returned limbs and the previously used ones.
   100  	clear(x.limbs[:max(n, len(x.limbs))])
   101  	x.limbs = x.limbs[:n]
   102  	return x
   103  }
   104  
   105  // resetToBytes assigns x = b, where b is a slice of big-endian bytes, resizing
   106  // n to the appropriate size.
   107  //
   108  // The announced length of x is set based on the actual bit size of the input,
   109  // ignoring leading zeroes.
   110  func (x *Nat) resetToBytes(b []byte) *Nat {
   111  	x.reset((len(b) + _S - 1) / _S)
   112  	if err := x.setBytes(b); err != nil {
   113  		panic("bigmod: internal error: bad arithmetic")
   114  	}
   115  	return x.trim()
   116  }
   117  
   118  // trim reduces the size of x to match its value.
   119  func (x *Nat) trim() *Nat {
   120  	// Trim most significant (trailing in little-endian) zero limbs.
   121  	// We assume comparison with zero (but not the branch) is constant time.
   122  	for i := len(x.limbs) - 1; i >= 0; i-- {
   123  		if x.limbs[i] != 0 {
   124  			break
   125  		}
   126  		x.limbs = x.limbs[:i]
   127  	}
   128  	return x
   129  }
   130  
   131  // set assigns x = y, optionally resizing x to the appropriate size.
   132  func (x *Nat) set(y *Nat) *Nat {
   133  	x.reset(len(y.limbs))
   134  	copy(x.limbs, y.limbs)
   135  	return x
   136  }
   137  
   138  // Bits returns x as a little-endian slice of uint. The length of the slice
   139  // matches the announced length of x. The result and x share the same underlying
   140  // array.
   141  func (x *Nat) Bits() []uint {
   142  	return x.limbs
   143  }
   144  
   145  // Bytes returns x as a zero-extended big-endian byte slice. The size of the
   146  // slice will match the size of m.
   147  //
   148  // x must have the same size as m and it must be less than or equal to m.
   149  func (x *Nat) Bytes(m *Modulus) []byte {
   150  	i := m.Size()
   151  	bytes := make([]byte, i)
   152  	for _, limb := range x.limbs {
   153  		for j := 0; j < _S; j++ {
   154  			i--
   155  			if i < 0 {
   156  				if limb == 0 {
   157  					break
   158  				}
   159  				panic("bigmod: modulus is smaller than nat")
   160  			}
   161  			bytes[i] = byte(limb)
   162  			limb >>= 8
   163  		}
   164  	}
   165  	return bytes
   166  }
   167  
   168  // SetBytes assigns x = b, where b is a slice of big-endian bytes.
   169  // SetBytes returns an error if b >= m.
   170  //
   171  // The output will be resized to the size of m and overwritten.
   172  //
   173  //go:norace
   174  func (x *Nat) SetBytes(b []byte, m *Modulus) (*Nat, error) {
   175  	x.resetFor(m)
   176  	if err := x.setBytes(b); err != nil {
   177  		return nil, err
   178  	}
   179  	if x.cmpGeq(m.nat) == yes {
   180  		return nil, errors.New("input overflows the modulus")
   181  	}
   182  	return x, nil
   183  }
   184  
   185  // SetOverflowingBytes assigns x = b, where b is a slice of big-endian bytes.
   186  // SetOverflowingBytes returns an error if b has a longer bit length than m, but
   187  // reduces overflowing values up to 2^⌈log2(m)⌉ - 1.
   188  //
   189  // The output will be resized to the size of m and overwritten.
   190  func (x *Nat) SetOverflowingBytes(b []byte, m *Modulus) (*Nat, error) {
   191  	x.resetFor(m)
   192  	if err := x.setBytes(b); err != nil {
   193  		return nil, err
   194  	}
   195  	// setBytes would have returned an error if the input overflowed the limb
   196  	// size of the modulus, so now we only need to check if the most significant
   197  	// limb of x has more bits than the most significant limb of the modulus.
   198  	if bitLen(x.limbs[len(x.limbs)-1]) > bitLen(m.nat.limbs[len(m.nat.limbs)-1]) {
   199  		return nil, errors.New("input overflows the modulus size")
   200  	}
   201  	x.maybeSubtractModulus(no, m)
   202  	return x, nil
   203  }
   204  
   205  // bigEndianUint returns the contents of buf interpreted as a
   206  // big-endian encoded uint value.
   207  func bigEndianUint(buf []byte) uint {
   208  	if _W == 64 {
   209  		return uint(byteorder.BEUint64(buf))
   210  	}
   211  	return uint(byteorder.BEUint32(buf))
   212  }
   213  
   214  func (x *Nat) setBytes(b []byte) error {
   215  	i, k := len(b), 0
   216  	for k < len(x.limbs) && i >= _S {
   217  		x.limbs[k] = bigEndianUint(b[i-_S : i])
   218  		i -= _S
   219  		k++
   220  	}
   221  	for s := 0; s < _W && k < len(x.limbs) && i > 0; s += 8 {
   222  		x.limbs[k] |= uint(b[i-1]) << s
   223  		i--
   224  	}
   225  	if i > 0 {
   226  		return errors.New("input overflows the modulus size")
   227  	}
   228  	return nil
   229  }
   230  
   231  // SetUint assigns x = y.
   232  //
   233  // The output will be resized to a single limb and overwritten.
   234  func (x *Nat) SetUint(y uint) *Nat {
   235  	x.reset(1)
   236  	x.limbs[0] = y
   237  	return x
   238  }
   239  
   240  // Equal returns 1 if x == y, and 0 otherwise.
   241  //
   242  // Both operands must have the same announced length.
   243  //
   244  //go:norace
   245  func (x *Nat) Equal(y *Nat) choice {
   246  	// Eliminate bounds checks in the loop.
   247  	size := len(x.limbs)
   248  	xLimbs := x.limbs[:size]
   249  	yLimbs := y.limbs[:size]
   250  
   251  	equal := yes
   252  	for i := 0; i < size; i++ {
   253  		equal &= ctEq(xLimbs[i], yLimbs[i])
   254  	}
   255  	return equal
   256  }
   257  
   258  // IsZero returns 1 if x == 0, and 0 otherwise.
   259  //
   260  //go:norace
   261  func (x *Nat) IsZero() choice {
   262  	// Eliminate bounds checks in the loop.
   263  	size := len(x.limbs)
   264  	xLimbs := x.limbs[:size]
   265  
   266  	zero := yes
   267  	for i := 0; i < size; i++ {
   268  		zero &= ctEq(xLimbs[i], 0)
   269  	}
   270  	return zero
   271  }
   272  
   273  // IsOne returns 1 if x == 1, and 0 otherwise.
   274  //
   275  //go:norace
   276  func (x *Nat) IsOne() choice {
   277  	// Eliminate bounds checks in the loop.
   278  	size := len(x.limbs)
   279  	xLimbs := x.limbs[:size]
   280  
   281  	if len(xLimbs) == 0 {
   282  		return no
   283  	}
   284  
   285  	one := ctEq(xLimbs[0], 1)
   286  	for i := 1; i < size; i++ {
   287  		one &= ctEq(xLimbs[i], 0)
   288  	}
   289  	return one
   290  }
   291  
   292  // IsMinusOne returns 1 if x == -1 mod m, and 0 otherwise.
   293  //
   294  // The length of x must be the same as the modulus. x must already be reduced
   295  // modulo m.
   296  //
   297  //go:norace
   298  func (x *Nat) IsMinusOne(m *Modulus) choice {
   299  	minusOne := m.Nat()
   300  	minusOne.SubOne(m)
   301  	return x.Equal(minusOne)
   302  }
   303  
   304  // IsOdd returns 1 if x is odd, and 0 otherwise.
   305  func (x *Nat) IsOdd() choice {
   306  	if len(x.limbs) == 0 {
   307  		return no
   308  	}
   309  	return choice(x.limbs[0] & 1)
   310  }
   311  
   312  // TrailingZeroBitsVarTime returns the number of trailing zero bits in x.
   313  func (x *Nat) TrailingZeroBitsVarTime() uint {
   314  	var t uint
   315  	limbs := x.limbs
   316  	for _, l := range limbs {
   317  		if l == 0 {
   318  			t += _W
   319  			continue
   320  		}
   321  		t += uint(bits.TrailingZeros(l))
   322  		break
   323  	}
   324  	return t
   325  }
   326  
   327  // cmpGeq returns 1 if x >= y, and 0 otherwise.
   328  //
   329  // Both operands must have the same announced length.
   330  //
   331  //go:norace
   332  func (x *Nat) cmpGeq(y *Nat) choice {
   333  	// Eliminate bounds checks in the loop.
   334  	size := len(x.limbs)
   335  	xLimbs := x.limbs[:size]
   336  	yLimbs := y.limbs[:size]
   337  
   338  	var c uint
   339  	for i := 0; i < size; i++ {
   340  		_, c = bits.Sub(xLimbs[i], yLimbs[i], c)
   341  	}
   342  	// If there was a carry, then subtracting y underflowed, so
   343  	// x is not greater than or equal to y.
   344  	return not(choice(c))
   345  }
   346  
   347  // assign sets x <- y if on == 1, and does nothing otherwise.
   348  //
   349  // Both operands must have the same announced length.
   350  //
   351  //go:norace
   352  func (x *Nat) assign(on choice, y *Nat) *Nat {
   353  	// Eliminate bounds checks in the loop.
   354  	size := len(x.limbs)
   355  	xLimbs := x.limbs[:size]
   356  	yLimbs := y.limbs[:size]
   357  
   358  	mask := ctMask(on)
   359  	for i := 0; i < size; i++ {
   360  		xLimbs[i] ^= mask & (xLimbs[i] ^ yLimbs[i])
   361  	}
   362  	return x
   363  }
   364  
   365  // add computes x += y and returns the carry.
   366  //
   367  // Both operands must have the same announced length.
   368  //
   369  //go:norace
   370  func (x *Nat) add(y *Nat) (c uint) {
   371  	// Eliminate bounds checks in the loop.
   372  	size := len(x.limbs)
   373  	xLimbs := x.limbs[:size]
   374  	yLimbs := y.limbs[:size]
   375  
   376  	for i := 0; i < size; i++ {
   377  		xLimbs[i], c = bits.Add(xLimbs[i], yLimbs[i], c)
   378  	}
   379  	return
   380  }
   381  
   382  // sub computes x -= y. It returns the borrow of the subtraction.
   383  //
   384  // Both operands must have the same announced length.
   385  //
   386  //go:norace
   387  func (x *Nat) sub(y *Nat) (c uint) {
   388  	// Eliminate bounds checks in the loop.
   389  	size := len(x.limbs)
   390  	xLimbs := x.limbs[:size]
   391  	yLimbs := y.limbs[:size]
   392  
   393  	for i := 0; i < size; i++ {
   394  		xLimbs[i], c = bits.Sub(xLimbs[i], yLimbs[i], c)
   395  	}
   396  	return
   397  }
   398  
   399  // ShiftRightVarTime sets x = x >> n.
   400  //
   401  // The announced length of x is unchanged.
   402  //
   403  //go:norace
   404  func (x *Nat) ShiftRightVarTime(n uint) *Nat {
   405  	// Eliminate bounds checks in the loop.
   406  	size := len(x.limbs)
   407  	xLimbs := x.limbs[:size]
   408  
   409  	shift := int(n % _W)
   410  	shiftLimbs := int(n / _W)
   411  
   412  	var shiftedLimbs []uint
   413  	if shiftLimbs < size {
   414  		shiftedLimbs = xLimbs[shiftLimbs:]
   415  	}
   416  
   417  	for i := range xLimbs {
   418  		if i >= len(shiftedLimbs) {
   419  			xLimbs[i] = 0
   420  			continue
   421  		}
   422  
   423  		xLimbs[i] = shiftedLimbs[i] >> shift
   424  		if i+1 < len(shiftedLimbs) {
   425  			xLimbs[i] |= shiftedLimbs[i+1] << (_W - shift)
   426  		}
   427  	}
   428  
   429  	return x
   430  }
   431  
   432  // BitLenVarTime returns the actual size of x in bits.
   433  //
   434  // The actual size of x (but nothing more) leaks through timing side-channels.
   435  // Note that this is ordinarily secret, as opposed to the announced size of x.
   436  func (x *Nat) BitLenVarTime() int {
   437  	// Eliminate bounds checks in the loop.
   438  	size := len(x.limbs)
   439  	xLimbs := x.limbs[:size]
   440  
   441  	for i := size - 1; i >= 0; i-- {
   442  		if xLimbs[i] != 0 {
   443  			return i*_W + bitLen(xLimbs[i])
   444  		}
   445  	}
   446  	return 0
   447  }
   448  
   449  // bitLen is a version of bits.Len that only leaks the bit length of n, but not
   450  // its value. bits.Len and bits.LeadingZeros use a lookup table for the
   451  // low-order bits on some architectures.
   452  func bitLen(n uint) int {
   453  	len := 0
   454  	// We assume, here and elsewhere, that comparison to zero is constant time
   455  	// with respect to different non-zero values.
   456  	for n != 0 {
   457  		len++
   458  		n >>= 1
   459  	}
   460  	return len
   461  }
   462  
   463  // Modulus is used for modular arithmetic, precomputing relevant constants.
   464  //
   465  // A Modulus can leak the exact number of bits needed to store its value
   466  // and is stored without padding. Its actual value is still kept secret.
   467  type Modulus struct {
   468  	// The underlying natural number for this modulus.
   469  	//
   470  	// This will be stored without any padding, and shouldn't alias with any
   471  	// other natural number being used.
   472  	nat *Nat
   473  
   474  	// If m is even, the following fields are not set.
   475  	odd   bool
   476  	m0inv uint // -nat.limbs[0]⁻¹ mod _W
   477  	rr    *Nat // R*R for montgomeryRepresentation
   478  }
   479  
   480  // rr returns R*R with R = 2^(_W * n) and n = len(m.nat.limbs).
   481  func rr(m *Modulus) *Nat {
   482  	rr := NewNat().ExpandFor(m)
   483  	n := uint(len(rr.limbs))
   484  	mLen := uint(m.BitLen())
   485  	logR := _W * n
   486  
   487  	// We start by computing R = 2^(_W * n) mod m. We can get pretty close, to
   488  	// 2^⌊log₂m⌋, by setting the highest bit we can without having to reduce.
   489  	rr.limbs[n-1] = 1 << ((mLen - 1) % _W)
   490  	// Then we double until we reach 2^(_W * n).
   491  	for i := mLen - 1; i < logR; i++ {
   492  		rr.Add(rr, m)
   493  	}
   494  
   495  	// Next we need to get from R to 2^(_W * n) R mod m (aka from one to R in
   496  	// the Montgomery domain, meaning we can use Montgomery multiplication now).
   497  	// We could do that by doubling _W * n times, or with a square-and-double
   498  	// chain log2(_W * n) long. Turns out the fastest thing is to start out with
   499  	// doublings, and switch to square-and-double once the exponent is large
   500  	// enough to justify the cost of the multiplications.
   501  
   502  	// The threshold is selected experimentally as a linear function of n.
   503  	threshold := n / 4
   504  
   505  	// We calculate how many of the most-significant bits of the exponent we can
   506  	// compute before crossing the threshold, and we do it with doublings.
   507  	i := bits.UintSize
   508  	for logR>>i <= threshold {
   509  		i--
   510  	}
   511  	for k := uint(0); k < logR>>i; k++ {
   512  		rr.Add(rr, m)
   513  	}
   514  
   515  	// Then we process the remaining bits of the exponent with a
   516  	// square-and-double chain.
   517  	for i > 0 {
   518  		rr.montgomeryMul(rr, rr, m)
   519  		i--
   520  		if logR>>i&1 != 0 {
   521  			rr.Add(rr, m)
   522  		}
   523  	}
   524  
   525  	return rr
   526  }
   527  
   528  // minusInverseModW computes -x⁻¹ mod _W with x odd.
   529  //
   530  // This operation is used to precompute a constant involved in Montgomery
   531  // multiplication.
   532  func minusInverseModW(x uint) uint {
   533  	// Every iteration of this loop doubles the least-significant bits of
   534  	// correct inverse in y. The first three bits are already correct (1⁻¹ = 1,
   535  	// 3⁻¹ = 3, 5⁻¹ = 5, and 7⁻¹ = 7 mod 8), so doubling five times is enough
   536  	// for 64 bits (and wastes only one iteration for 32 bits).
   537  	//
   538  	// See https://crypto.stackexchange.com/a/47496.
   539  	y := x
   540  	for i := 0; i < 5; i++ {
   541  		y = y * (2 - x*y)
   542  	}
   543  	return -y
   544  }
   545  
   546  // NewModulus creates a new Modulus from a slice of big-endian bytes. The
   547  // modulus must be greater than one.
   548  //
   549  // The number of significant bits and whether the modulus is even is leaked
   550  // through timing side-channels.
   551  func NewModulus(b []byte) (*Modulus, error) {
   552  	n := NewNat().resetToBytes(b)
   553  	return newModulus(n)
   554  }
   555  
   556  // NewModulusProduct creates a new Modulus from the product of two numbers
   557  // represented as big-endian byte slices. The result must be greater than one.
   558  //
   559  //go:norace
   560  func NewModulusProduct(a, b []byte) (*Modulus, error) {
   561  	x := NewNat().resetToBytes(a)
   562  	y := NewNat().resetToBytes(b)
   563  	n := NewNat().reset(len(x.limbs) + len(y.limbs))
   564  	for i := range y.limbs {
   565  		n.limbs[i+len(x.limbs)] = addMulVVW(n.limbs[i:i+len(x.limbs)], x.limbs, y.limbs[i])
   566  	}
   567  	return newModulus(n.trim())
   568  }
   569  
   570  func newModulus(n *Nat) (*Modulus, error) {
   571  	m := &Modulus{nat: n}
   572  	if m.nat.IsZero() == yes || m.nat.IsOne() == yes {
   573  		return nil, errors.New("modulus must be > 1")
   574  	}
   575  	if m.nat.IsOdd() == 1 {
   576  		m.odd = true
   577  		m.m0inv = minusInverseModW(m.nat.limbs[0])
   578  		m.rr = rr(m)
   579  	}
   580  	return m, nil
   581  }
   582  
   583  // Size returns the size of m in bytes.
   584  func (m *Modulus) Size() int {
   585  	return (m.BitLen() + 7) / 8
   586  }
   587  
   588  // BitLen returns the size of m in bits.
   589  func (m *Modulus) BitLen() int {
   590  	return m.nat.BitLenVarTime()
   591  }
   592  
   593  // Nat returns m as a Nat.
   594  func (m *Modulus) Nat() *Nat {
   595  	// Make a copy so that the caller can't modify m.nat or alias it with
   596  	// another Nat in a modulus operation.
   597  	n := NewNat()
   598  	n.set(m.nat)
   599  	return n
   600  }
   601  
   602  // shiftIn calculates x = x << _W + y mod m.
   603  //
   604  // This assumes that x is already reduced mod m.
   605  //
   606  //go:norace
   607  func (x *Nat) shiftIn(y uint, m *Modulus) *Nat {
   608  	d := NewNat().resetFor(m)
   609  
   610  	// Eliminate bounds checks in the loop.
   611  	size := len(m.nat.limbs)
   612  	xLimbs := x.limbs[:size]
   613  	dLimbs := d.limbs[:size]
   614  	mLimbs := m.nat.limbs[:size]
   615  
   616  	// Each iteration of this loop computes x = 2x + b mod m, where b is a bit
   617  	// from y. Effectively, it left-shifts x and adds y one bit at a time,
   618  	// reducing it every time.
   619  	//
   620  	// To do the reduction, each iteration computes both 2x + b and 2x + b - m.
   621  	// The next iteration (and finally the return line) will use either result
   622  	// based on whether 2x + b overflows m.
   623  	needSubtraction := no
   624  	for i := _W - 1; i >= 0; i-- {
   625  		carry := (y >> i) & 1
   626  		var borrow uint
   627  		mask := ctMask(needSubtraction)
   628  		for i := 0; i < size; i++ {
   629  			l := xLimbs[i] ^ (mask & (xLimbs[i] ^ dLimbs[i]))
   630  			xLimbs[i], carry = bits.Add(l, l, carry)
   631  			dLimbs[i], borrow = bits.Sub(xLimbs[i], mLimbs[i], borrow)
   632  		}
   633  		// Like in maybeSubtractModulus, we need the subtraction if either it
   634  		// didn't underflow (meaning 2x + b > m) or if computing 2x + b
   635  		// overflowed (meaning 2x + b > 2^_W*n > m).
   636  		needSubtraction = not(choice(borrow)) | choice(carry)
   637  	}
   638  	return x.assign(needSubtraction, d)
   639  }
   640  
   641  // Mod calculates out = x mod m.
   642  //
   643  // This works regardless how large the value of x is.
   644  //
   645  // The output will be resized to the size of m and overwritten.
   646  //
   647  //go:norace
   648  func (out *Nat) Mod(x *Nat, m *Modulus) *Nat {
   649  	out.resetFor(m)
   650  	// Working our way from the most significant to the least significant limb,
   651  	// we can insert each limb at the least significant position, shifting all
   652  	// previous limbs left by _W. This way each limb will get shifted by the
   653  	// correct number of bits. We can insert at least N - 1 limbs without
   654  	// overflowing m. After that, we need to reduce every time we shift.
   655  	i := len(x.limbs) - 1
   656  	// For the first N - 1 limbs we can skip the actual shifting and position
   657  	// them at the shifted position, which starts at min(N - 2, i).
   658  	start := len(m.nat.limbs) - 2
   659  	if i < start {
   660  		start = i
   661  	}
   662  	for j := start; j >= 0; j-- {
   663  		out.limbs[j] = x.limbs[i]
   664  		i--
   665  	}
   666  	// We shift in the remaining limbs, reducing modulo m each time.
   667  	for i >= 0 {
   668  		out.shiftIn(x.limbs[i], m)
   669  		i--
   670  	}
   671  	return out
   672  }
   673  
   674  // ExpandFor ensures x has the right size to work with operations modulo m.
   675  //
   676  // The announced size of x must be smaller than or equal to that of m.
   677  func (x *Nat) ExpandFor(m *Modulus) *Nat {
   678  	return x.expand(len(m.nat.limbs))
   679  }
   680  
   681  // resetFor ensures out has the right size to work with operations modulo m.
   682  //
   683  // out is zeroed and may start at any size.
   684  func (out *Nat) resetFor(m *Modulus) *Nat {
   685  	return out.reset(len(m.nat.limbs))
   686  }
   687  
   688  // maybeSubtractModulus computes x -= m if and only if x >= m or if "always" is yes.
   689  //
   690  // It can be used to reduce modulo m a value up to 2m - 1, which is a common
   691  // range for results computed by higher level operations.
   692  //
   693  // always is usually a carry that indicates that the operation that produced x
   694  // overflowed its size, meaning abstractly x > 2^_W*n > m even if x < m.
   695  //
   696  // x and m operands must have the same announced length.
   697  //
   698  //go:norace
   699  func (x *Nat) maybeSubtractModulus(always choice, m *Modulus) {
   700  	t := NewNat().set(x)
   701  	underflow := t.sub(m.nat)
   702  	// We keep the result if x - m didn't underflow (meaning x >= m)
   703  	// or if always was set.
   704  	keep := not(choice(underflow)) | choice(always)
   705  	x.assign(keep, t)
   706  }
   707  
   708  // Sub computes x = x - y mod m.
   709  //
   710  // The length of both operands must be the same as the modulus. Both operands
   711  // must already be reduced modulo m.
   712  //
   713  //go:norace
   714  func (x *Nat) Sub(y *Nat, m *Modulus) *Nat {
   715  	underflow := x.sub(y)
   716  	// If the subtraction underflowed, add m.
   717  	t := NewNat().set(x)
   718  	t.add(m.nat)
   719  	x.assign(choice(underflow), t)
   720  	return x
   721  }
   722  
   723  // SubOne computes x = x - 1 mod m.
   724  //
   725  // The length of x must be the same as the modulus.
   726  func (x *Nat) SubOne(m *Modulus) *Nat {
   727  	one := NewNat().ExpandFor(m)
   728  	one.limbs[0] = 1
   729  	// Sub asks for x to be reduced modulo m, while SubOne doesn't, but when
   730  	// y = 1, it works, and this is an internal use.
   731  	return x.Sub(one, m)
   732  }
   733  
   734  // Add computes x = x + y mod m.
   735  //
   736  // The length of both operands must be the same as the modulus. Both operands
   737  // must already be reduced modulo m.
   738  //
   739  //go:norace
   740  func (x *Nat) Add(y *Nat, m *Modulus) *Nat {
   741  	overflow := x.add(y)
   742  	x.maybeSubtractModulus(choice(overflow), m)
   743  	return x
   744  }
   745  
   746  // montgomeryRepresentation calculates x = x * R mod m, with R = 2^(_W * n) and
   747  // n = len(m.nat.limbs).
   748  //
   749  // Faster Montgomery multiplication replaces standard modular multiplication for
   750  // numbers in this representation.
   751  //
   752  // This assumes that x is already reduced mod m.
   753  func (x *Nat) montgomeryRepresentation(m *Modulus) *Nat {
   754  	// A Montgomery multiplication (which computes a * b / R) by R * R works out
   755  	// to a multiplication by R, which takes the value out of the Montgomery domain.
   756  	return x.montgomeryMul(x, m.rr, m)
   757  }
   758  
   759  // montgomeryReduction calculates x = x / R mod m, with R = 2^(_W * n) and
   760  // n = len(m.nat.limbs).
   761  //
   762  // This assumes that x is already reduced mod m.
   763  func (x *Nat) montgomeryReduction(m *Modulus) *Nat {
   764  	// By Montgomery multiplying with 1 not in Montgomery representation, we
   765  	// convert out back from Montgomery representation, because it works out to
   766  	// dividing by R.
   767  	one := NewNat().ExpandFor(m)
   768  	one.limbs[0] = 1
   769  	return x.montgomeryMul(x, one, m)
   770  }
   771  
   772  // montgomeryMul calculates x = a * b / R mod m, with R = 2^(_W * n) and
   773  // n = len(m.nat.limbs), also known as a Montgomery multiplication.
   774  //
   775  // All inputs should be the same length and already reduced modulo m.
   776  // x will be resized to the size of m and overwritten.
   777  //
   778  //go:norace
   779  func (x *Nat) montgomeryMul(a *Nat, b *Nat, m *Modulus) *Nat {
   780  	n := len(m.nat.limbs)
   781  	mLimbs := m.nat.limbs[:n]
   782  	aLimbs := a.limbs[:n]
   783  	bLimbs := b.limbs[:n]
   784  
   785  	switch n {
   786  	default:
   787  		// Attempt to use a stack-allocated backing array.
   788  		T := make([]uint, 0, preallocLimbs*2)
   789  		if cap(T) < n*2 {
   790  			T = make([]uint, 0, n*2)
   791  		}
   792  		T = T[:n*2]
   793  
   794  		// This loop implements Word-by-Word Montgomery Multiplication, as
   795  		// described in Algorithm 4 (Fig. 3) of "Efficient Software
   796  		// Implementations of Modular Exponentiation" by Shay Gueron
   797  		// [https://eprint.iacr.org/2011/239.pdf].
   798  		var c uint
   799  		for i := 0; i < n; i++ {
   800  			_ = T[n+i] // bounds check elimination hint
   801  
   802  			// Step 1 (T = a × b) is computed as a large pen-and-paper column
   803  			// multiplication of two numbers with n base-2^_W digits. If we just
   804  			// wanted to produce 2n-wide T, we would do
   805  			//
   806  			//   for i := 0; i < n; i++ {
   807  			//       d := bLimbs[i]
   808  			//       T[n+i] = addMulVVW(T[i:n+i], aLimbs, d)
   809  			//   }
   810  			//
   811  			// where d is a digit of the multiplier, T[i:n+i] is the shifted
   812  			// position of the product of that digit, and T[n+i] is the final carry.
   813  			// Note that T[i] isn't modified after processing the i-th digit.
   814  			//
   815  			// Instead of running two loops, one for Step 1 and one for Steps 2–6,
   816  			// the result of Step 1 is computed during the next loop. This is
   817  			// possible because each iteration only uses T[i] in Step 2 and then
   818  			// discards it in Step 6.
   819  			d := bLimbs[i]
   820  			c1 := addMulVVW(T[i:n+i], aLimbs, d)
   821  
   822  			// Step 6 is replaced by shifting the virtual window we operate
   823  			// over: T of the algorithm is T[i:] for us. That means that T1 in
   824  			// Step 2 (T mod 2^_W) is simply T[i]. k0 in Step 3 is our m0inv.
   825  			Y := T[i] * m.m0inv
   826  
   827  			// Step 4 and 5 add Y × m to T, which as mentioned above is stored
   828  			// at T[i:]. The two carries (from a × d and Y × m) are added up in
   829  			// the next word T[n+i], and the carry bit from that addition is
   830  			// brought forward to the next iteration.
   831  			c2 := addMulVVW(T[i:n+i], mLimbs, Y)
   832  			T[n+i], c = bits.Add(c1, c2, c)
   833  		}
   834  
   835  		// Finally for Step 7 we copy the final T window into x, and subtract m
   836  		// if necessary (which as explained in maybeSubtractModulus can be the
   837  		// case both if x >= m, or if x overflowed).
   838  		//
   839  		// The paper suggests in Section 4 that we can do an "Almost Montgomery
   840  		// Multiplication" by subtracting only in the overflow case, but the
   841  		// cost is very similar since the constant time subtraction tells us if
   842  		// x >= m as a side effect, and taking care of the broken invariant is
   843  		// highly undesirable (see https://go.dev/issue/13907).
   844  		copy(x.reset(n).limbs, T[n:])
   845  		x.maybeSubtractModulus(choice(c), m)
   846  
   847  	// The following specialized cases follow the exact same algorithm, but
   848  	// optimized for the sizes most used in RSA. addMulVVW is implemented in
   849  	// assembly with loop unrolling depending on the architecture and bounds
   850  	// checks are removed by the compiler thanks to the constant size.
   851  	case 1024 / _W:
   852  		const n = 1024 / _W // compiler hint
   853  		T := make([]uint, n*2)
   854  		var c uint
   855  		for i := 0; i < n; i++ {
   856  			d := bLimbs[i]
   857  			c1 := addMulVVW1024(&T[i], &aLimbs[0], d)
   858  			Y := T[i] * m.m0inv
   859  			c2 := addMulVVW1024(&T[i], &mLimbs[0], Y)
   860  			T[n+i], c = bits.Add(c1, c2, c)
   861  		}
   862  		copy(x.reset(n).limbs, T[n:])
   863  		x.maybeSubtractModulus(choice(c), m)
   864  
   865  	case 1536 / _W:
   866  		const n = 1536 / _W // compiler hint
   867  		T := make([]uint, n*2)
   868  		var c uint
   869  		for i := 0; i < n; i++ {
   870  			d := bLimbs[i]
   871  			c1 := addMulVVW1536(&T[i], &aLimbs[0], d)
   872  			Y := T[i] * m.m0inv
   873  			c2 := addMulVVW1536(&T[i], &mLimbs[0], Y)
   874  			T[n+i], c = bits.Add(c1, c2, c)
   875  		}
   876  		copy(x.reset(n).limbs, T[n:])
   877  		x.maybeSubtractModulus(choice(c), m)
   878  
   879  	case 2048 / _W:
   880  		const n = 2048 / _W // compiler hint
   881  		T := make([]uint, n*2)
   882  		var c uint
   883  		for i := 0; i < n; i++ {
   884  			d := bLimbs[i]
   885  			c1 := addMulVVW2048(&T[i], &aLimbs[0], d)
   886  			Y := T[i] * m.m0inv
   887  			c2 := addMulVVW2048(&T[i], &mLimbs[0], Y)
   888  			T[n+i], c = bits.Add(c1, c2, c)
   889  		}
   890  		copy(x.reset(n).limbs, T[n:])
   891  		x.maybeSubtractModulus(choice(c), m)
   892  	}
   893  
   894  	return x
   895  }
   896  
   897  // addMulVVW multiplies the multi-word value x by the single-word value y,
   898  // adding the result to the multi-word value z and returning the final carry.
   899  // It can be thought of as one row of a pen-and-paper column multiplication.
   900  //
   901  //go:norace
   902  func addMulVVW(z, x []uint, y uint) (carry uint) {
   903  	_ = x[len(z)-1] // bounds check elimination hint
   904  	for i := range z {
   905  		hi, lo := bits.Mul(x[i], y)
   906  		lo, c := bits.Add(lo, z[i], 0)
   907  		// We use bits.Add with zero to get an add-with-carry instruction that
   908  		// absorbs the carry from the previous bits.Add.
   909  		hi, _ = bits.Add(hi, 0, c)
   910  		lo, c = bits.Add(lo, carry, 0)
   911  		hi, _ = bits.Add(hi, 0, c)
   912  		carry = hi
   913  		z[i] = lo
   914  	}
   915  	return carry
   916  }
   917  
   918  // Mul calculates x = x * y mod m.
   919  //
   920  // The length of both operands must be the same as the modulus. Both operands
   921  // must already be reduced modulo m.
   922  //
   923  //go:norace
   924  func (x *Nat) Mul(y *Nat, m *Modulus) *Nat {
   925  	if m.odd {
   926  		// A Montgomery multiplication by a value out of the Montgomery domain
   927  		// takes the result out of Montgomery representation.
   928  		xR := NewNat().set(x).montgomeryRepresentation(m) // xR = x * R mod m
   929  		return x.montgomeryMul(xR, y, m)                  // x = xR * y / R mod m
   930  	}
   931  
   932  	n := len(m.nat.limbs)
   933  	xLimbs := x.limbs[:n]
   934  	yLimbs := y.limbs[:n]
   935  
   936  	switch n {
   937  	default:
   938  		// Attempt to use a stack-allocated backing array.
   939  		T := make([]uint, 0, preallocLimbs*2)
   940  		if cap(T) < n*2 {
   941  			T = make([]uint, 0, n*2)
   942  		}
   943  		T = T[:n*2]
   944  
   945  		// T = x * y
   946  		for i := 0; i < n; i++ {
   947  			T[n+i] = addMulVVW(T[i:n+i], xLimbs, yLimbs[i])
   948  		}
   949  
   950  		// x = T mod m
   951  		return x.Mod(&Nat{limbs: T}, m)
   952  
   953  	// The following specialized cases follow the exact same algorithm, but
   954  	// optimized for the sizes most used in RSA. See montgomeryMul for details.
   955  	case 1024 / _W:
   956  		const n = 1024 / _W // compiler hint
   957  		T := make([]uint, n*2)
   958  		for i := 0; i < n; i++ {
   959  			T[n+i] = addMulVVW1024(&T[i], &xLimbs[0], yLimbs[i])
   960  		}
   961  		return x.Mod(&Nat{limbs: T}, m)
   962  	case 1536 / _W:
   963  		const n = 1536 / _W // compiler hint
   964  		T := make([]uint, n*2)
   965  		for i := 0; i < n; i++ {
   966  			T[n+i] = addMulVVW1536(&T[i], &xLimbs[0], yLimbs[i])
   967  		}
   968  		return x.Mod(&Nat{limbs: T}, m)
   969  	case 2048 / _W:
   970  		const n = 2048 / _W // compiler hint
   971  		T := make([]uint, n*2)
   972  		for i := 0; i < n; i++ {
   973  			T[n+i] = addMulVVW2048(&T[i], &xLimbs[0], yLimbs[i])
   974  		}
   975  		return x.Mod(&Nat{limbs: T}, m)
   976  	}
   977  }
   978  
   979  // Exp calculates out = x^e mod m.
   980  //
   981  // The exponent e is represented in big-endian order. The output will be resized
   982  // to the size of m and overwritten. x must already be reduced modulo m.
   983  //
   984  // m must be odd, or Exp will panic.
   985  //
   986  //go:norace
   987  func (out *Nat) Exp(x *Nat, e []byte, m *Modulus) *Nat {
   988  	if !m.odd {
   989  		panic("bigmod: modulus for Exp must be odd")
   990  	}
   991  
   992  	// We use a 4 bit window. For our RSA workload, 4 bit windows are faster
   993  	// than 2 bit windows, but use an extra 12 nats worth of scratch space.
   994  	// Using bit sizes that don't divide 8 are more complex to implement, but
   995  	// are likely to be more efficient if necessary.
   996  
   997  	table := [(1 << 4) - 1]*Nat{ // table[i] = x ^ (i+1)
   998  		// newNat calls are unrolled so they are allocated on the stack.
   999  		NewNat(), NewNat(), NewNat(), NewNat(), NewNat(),
  1000  		NewNat(), NewNat(), NewNat(), NewNat(), NewNat(),
  1001  		NewNat(), NewNat(), NewNat(), NewNat(), NewNat(),
  1002  	}
  1003  	table[0].set(x).montgomeryRepresentation(m)
  1004  	for i := 1; i < len(table); i++ {
  1005  		table[i].montgomeryMul(table[i-1], table[0], m)
  1006  	}
  1007  
  1008  	out.resetFor(m)
  1009  	out.limbs[0] = 1
  1010  	out.montgomeryRepresentation(m)
  1011  	tmp := NewNat().ExpandFor(m)
  1012  	for _, b := range e {
  1013  		for _, j := range []int{4, 0} {
  1014  			// Square four times. Optimization note: this can be implemented
  1015  			// more efficiently than with generic Montgomery multiplication.
  1016  			out.montgomeryMul(out, out, m)
  1017  			out.montgomeryMul(out, out, m)
  1018  			out.montgomeryMul(out, out, m)
  1019  			out.montgomeryMul(out, out, m)
  1020  
  1021  			// Select x^k in constant time from the table.
  1022  			k := uint((b >> j) & 0b1111)
  1023  			for i := range table {
  1024  				tmp.assign(ctEq(k, uint(i+1)), table[i])
  1025  			}
  1026  
  1027  			// Multiply by x^k, discarding the result if k = 0.
  1028  			tmp.montgomeryMul(out, tmp, m)
  1029  			out.assign(not(ctEq(k, 0)), tmp)
  1030  		}
  1031  	}
  1032  
  1033  	return out.montgomeryReduction(m)
  1034  }
  1035  
  1036  // ExpShortVarTime calculates out = x^e mod m.
  1037  //
  1038  // The output will be resized to the size of m and overwritten. x must already
  1039  // be reduced modulo m. This leaks the exponent through timing side-channels.
  1040  //
  1041  // m must be odd, or ExpShortVarTime will panic.
  1042  func (out *Nat) ExpShortVarTime(x *Nat, e uint, m *Modulus) *Nat {
  1043  	if !m.odd {
  1044  		panic("bigmod: modulus for ExpShortVarTime must be odd")
  1045  	}
  1046  	// For short exponents, precomputing a table and using a window like in Exp
  1047  	// doesn't pay off. Instead, we do a simple conditional square-and-multiply
  1048  	// chain, skipping the initial run of zeroes.
  1049  	xR := NewNat().set(x).montgomeryRepresentation(m)
  1050  	out.set(xR)
  1051  	for i := bits.UintSize - bits.Len(e) + 1; i < bits.UintSize; i++ {
  1052  		out.montgomeryMul(out, out, m)
  1053  		if k := (e >> (bits.UintSize - i - 1)) & 1; k != 0 {
  1054  			out.montgomeryMul(out, xR, m)
  1055  		}
  1056  	}
  1057  	return out.montgomeryReduction(m)
  1058  }
  1059  
  1060  // InverseVarTime calculates x = a⁻¹ mod m and returns (x, true) if a is
  1061  // invertible. Otherwise, InverseVarTime returns (x, false) and x is not
  1062  // modified.
  1063  //
  1064  // a must be reduced modulo m, but doesn't need to have the same size. The
  1065  // output will be resized to the size of m and overwritten.
  1066  //
  1067  //go:norace
  1068  func (x *Nat) InverseVarTime(a *Nat, m *Modulus) (*Nat, bool) {
  1069  	u, A, err := extendedGCD(a, m.nat)
  1070  	if err != nil {
  1071  		return x, false
  1072  	}
  1073  	if u.IsOne() == no {
  1074  		return x, false
  1075  	}
  1076  	return x.set(A), true
  1077  }
  1078  
  1079  // GCDVarTime calculates x = GCD(a, b) where at least one of a or b is odd, and
  1080  // both are non-zero. If GCDVarTime returns an error, x is not modified.
  1081  //
  1082  // The output will be resized to the size of the larger of a and b.
  1083  func (x *Nat) GCDVarTime(a, b *Nat) (*Nat, error) {
  1084  	u, _, err := extendedGCD(a, b)
  1085  	if err != nil {
  1086  		return nil, err
  1087  	}
  1088  	return x.set(u), nil
  1089  }
  1090  
  1091  // extendedGCD computes u and A such that a = GCD(a, m) and u = A*a - B*m.
  1092  //
  1093  // u will have the size of the larger of a and m, and A will have the size of m.
  1094  //
  1095  // It is an error if either a or m is zero, or if they are both even.
  1096  func extendedGCD(a, m *Nat) (u, A *Nat, err error) {
  1097  	// This is the extended binary GCD algorithm described in the Handbook of
  1098  	// Applied Cryptography, Algorithm 14.61, adapted by BoringSSL to bound
  1099  	// coefficients and avoid negative numbers. For more details and proof of
  1100  	// correctness, see https://github.com/mit-plv/fiat-crypto/pull/333/files.
  1101  	//
  1102  	// Following the proof linked in the PR above, the changes are:
  1103  	//
  1104  	// 1. Negate [B] and [C] so they are positive. The invariant now involves a
  1105  	//    subtraction.
  1106  	// 2. If step 2 (both [x] and [y] are even) runs, abort immediately. This
  1107  	//    case needs to be handled by the caller.
  1108  	// 3. Subtract copies of [x] and [y] as needed in step 6 (both [u] and [v]
  1109  	//    are odd) so coefficients stay in bounds.
  1110  	// 4. Replace the [u >= v] check with [u > v]. This changes the end
  1111  	//    condition to [v = 0] rather than [u = 0]. This saves an extra
  1112  	//    subtraction due to which coefficients were negated.
  1113  	// 5. Rename x and y to a and n, to capture that one is a modulus.
  1114  	// 6. Rearrange steps 4 through 6 slightly. Merge the loops in steps 4 and
  1115  	//    5 into the main loop (step 7's goto), and move step 6 to the start of
  1116  	//    the loop iteration, ensuring each loop iteration halves at least one
  1117  	//    value.
  1118  	//
  1119  	// Note this algorithm does not handle either input being zero.
  1120  
  1121  	if a.IsZero() == yes || m.IsZero() == yes {
  1122  		return nil, nil, errors.New("extendedGCD: a or m is zero")
  1123  	}
  1124  	if a.IsOdd() == no && m.IsOdd() == no {
  1125  		return nil, nil, errors.New("extendedGCD: both a and m are even")
  1126  	}
  1127  
  1128  	size := max(len(a.limbs), len(m.limbs))
  1129  	u = NewNat().set(a).expand(size)
  1130  	v := NewNat().set(m).expand(size)
  1131  
  1132  	A = NewNat().reset(len(m.limbs))
  1133  	A.limbs[0] = 1
  1134  	B := NewNat().reset(len(a.limbs))
  1135  	C := NewNat().reset(len(m.limbs))
  1136  	D := NewNat().reset(len(a.limbs))
  1137  	D.limbs[0] = 1
  1138  
  1139  	// Before and after each loop iteration, the following hold:
  1140  	//
  1141  	//   u = A*a - B*m
  1142  	//   v = D*m - C*a
  1143  	//   0 < u <= a
  1144  	//   0 <= v <= m
  1145  	//   0 <= A < m
  1146  	//   0 <= B <= a
  1147  	//   0 <= C < m
  1148  	//   0 <= D <= a
  1149  	//
  1150  	// After each loop iteration, u and v only get smaller, and at least one of
  1151  	// them shrinks by at least a factor of two.
  1152  	for {
  1153  		// If both u and v are odd, subtract the smaller from the larger.
  1154  		// If u = v, we need to subtract from v to hit the modified exit condition.
  1155  		if u.IsOdd() == yes && v.IsOdd() == yes {
  1156  			if v.cmpGeq(u) == no {
  1157  				u.sub(v)
  1158  				A.Add(C, &Modulus{nat: m})
  1159  				B.Add(D, &Modulus{nat: a})
  1160  			} else {
  1161  				v.sub(u)
  1162  				C.Add(A, &Modulus{nat: m})
  1163  				D.Add(B, &Modulus{nat: a})
  1164  			}
  1165  		}
  1166  
  1167  		// Exactly one of u and v is now even.
  1168  		if u.IsOdd() == v.IsOdd() {
  1169  			panic("bigmod: internal error: u and v are not in the expected state")
  1170  		}
  1171  
  1172  		// Halve the even one and adjust the corresponding coefficient.
  1173  		if u.IsOdd() == no {
  1174  			rshift1(u, 0)
  1175  			if A.IsOdd() == yes || B.IsOdd() == yes {
  1176  				rshift1(A, A.add(m))
  1177  				rshift1(B, B.add(a))
  1178  			} else {
  1179  				rshift1(A, 0)
  1180  				rshift1(B, 0)
  1181  			}
  1182  		} else { // v.IsOdd() == no
  1183  			rshift1(v, 0)
  1184  			if C.IsOdd() == yes || D.IsOdd() == yes {
  1185  				rshift1(C, C.add(m))
  1186  				rshift1(D, D.add(a))
  1187  			} else {
  1188  				rshift1(C, 0)
  1189  				rshift1(D, 0)
  1190  			}
  1191  		}
  1192  
  1193  		if v.IsZero() == yes {
  1194  			return u, A, nil
  1195  		}
  1196  	}
  1197  }
  1198  
  1199  //go:norace
  1200  func rshift1(a *Nat, carry uint) {
  1201  	size := len(a.limbs)
  1202  	aLimbs := a.limbs[:size]
  1203  
  1204  	for i := range size {
  1205  		aLimbs[i] >>= 1
  1206  		if i+1 < size {
  1207  			aLimbs[i] |= aLimbs[i+1] << (_W - 1)
  1208  		} else {
  1209  			aLimbs[i] |= carry << (_W - 1)
  1210  		}
  1211  	}
  1212  }
  1213  
  1214  // DivShortVarTime calculates x = x / y and returns the remainder.
  1215  //
  1216  // It panics if y is zero.
  1217  //
  1218  //go:norace
  1219  func (x *Nat) DivShortVarTime(y uint) uint {
  1220  	if y == 0 {
  1221  		panic("bigmod: division by zero")
  1222  	}
  1223  
  1224  	var r uint
  1225  	for i := len(x.limbs) - 1; i >= 0; i-- {
  1226  		x.limbs[i], r = bits.Div(r, x.limbs[i], y)
  1227  	}
  1228  	return r
  1229  }
  1230  

View as plain text