mirror of
https://github.com/vlang/v.git
synced 2023-08-10 21:13:21 +03:00
465 lines
12 KiB
V
465 lines
12 KiB
V
module edwards25519
|
|
|
|
import os
|
|
import rand
|
|
import math.bits
|
|
import math.big
|
|
import encoding.hex
|
|
|
|
const github_job = os.getenv('GITHUB_JOB')
|
|
|
|
fn testsuite_begin() {
|
|
if edwards25519.github_job != '' {
|
|
// ensure that the CI does not run flaky tests:
|
|
rand.seed([u32(0xffff24), 0xabcd])
|
|
}
|
|
}
|
|
|
|
fn (mut v Element) str() string {
|
|
return hex.encode(v.bytes())
|
|
}
|
|
|
|
const mask_low_52_bits = (u64(1) << 52) - 1
|
|
|
|
fn generate_field_element() Element {
|
|
return Element{
|
|
l0: rand.u64() & edwards25519.mask_low_52_bits
|
|
l1: rand.u64() & edwards25519.mask_low_52_bits
|
|
l2: rand.u64() & edwards25519.mask_low_52_bits
|
|
l3: rand.u64() & edwards25519.mask_low_52_bits
|
|
l4: rand.u64() & edwards25519.mask_low_52_bits
|
|
}
|
|
}
|
|
|
|
// weirdLimbs can be combined to generate a range of edge-case edwards25519 elements.
|
|
// 0 and -1 are intentionally more weighted, as they combine well.
|
|
const (
|
|
two_to_51 = u64(1) << 51
|
|
two_to_52 = u64(1) << 52
|
|
weird_limbs_51 = [
|
|
u64(0),
|
|
0,
|
|
0,
|
|
0,
|
|
1,
|
|
19 - 1,
|
|
19,
|
|
0x2aaaaaaaaaaaa,
|
|
0x5555555555555,
|
|
two_to_51 - 20,
|
|
two_to_51 - 19,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
]
|
|
weird_limbs_52 = [
|
|
u64(0),
|
|
0,
|
|
0,
|
|
0,
|
|
0,
|
|
0,
|
|
1,
|
|
19 - 1,
|
|
19,
|
|
0x2aaaaaaaaaaaa,
|
|
0x5555555555555,
|
|
two_to_51 - 20,
|
|
two_to_51 - 19,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51 - 1,
|
|
two_to_51,
|
|
two_to_51 + 1,
|
|
two_to_52 - 19,
|
|
two_to_52 - 1,
|
|
]
|
|
)
|
|
|
|
fn generate_weird_field_element() Element {
|
|
return Element{
|
|
l0: edwards25519.weird_limbs_52[rand.intn(edwards25519.weird_limbs_52.len) or { 0 }]
|
|
l1: edwards25519.weird_limbs_51[rand.intn(edwards25519.weird_limbs_51.len) or { 0 }]
|
|
l2: edwards25519.weird_limbs_51[rand.intn(edwards25519.weird_limbs_51.len) or { 0 }]
|
|
l3: edwards25519.weird_limbs_51[rand.intn(edwards25519.weird_limbs_51.len) or { 0 }]
|
|
l4: edwards25519.weird_limbs_51[rand.intn(edwards25519.weird_limbs_51.len) or { 0 }]
|
|
}
|
|
}
|
|
|
|
fn (e Element) generate_element() Element {
|
|
if rand.intn(2) or { 0 } == 0 {
|
|
return generate_weird_field_element()
|
|
}
|
|
return generate_field_element()
|
|
}
|
|
|
|
fn is_in_bounds(x Element) bool {
|
|
return bits.len_64(x.l0) <= 52 && bits.len_64(x.l1) <= 52 && bits.len_64(x.l2) <= 52
|
|
&& bits.len_64(x.l3) <= 52 && bits.len_64(x.l4) <= 52
|
|
}
|
|
|
|
fn carry_gen(a [5]u64) bool {
|
|
mut t1 := Element{a[0], a[1], a[2], a[3], a[4]}
|
|
mut t2 := Element{a[0], a[1], a[2], a[3], a[4]}
|
|
|
|
t1.carry_propagate_generic()
|
|
t2.carry_propagate_generic()
|
|
|
|
return t1 == t2 && is_in_bounds(t2)
|
|
}
|
|
|
|
fn test_carry_propagate_generic() {
|
|
// closures not supported on windows
|
|
for i := 0; i <= 10; i++ {
|
|
els := [rand.u64(), rand.u64(), rand.u64(), rand.u64(),
|
|
rand.u64()]!
|
|
p := carry_gen(els)
|
|
assert p == true
|
|
}
|
|
res := carry_gen([u64(0xffffffffffffffff), 0xffffffffffffffff, 0xffffffffffffffff,
|
|
0xffffffffffffffff, 0xffffffffffffffff]!)
|
|
assert res == true
|
|
}
|
|
|
|
fn test_fe_mul_generic() {
|
|
for i in 0 .. 20 {
|
|
el := Element{}
|
|
a := el.generate_element()
|
|
b := el.generate_element()
|
|
a1 := a
|
|
a2 := a
|
|
|
|
b1 := b
|
|
b2 := b
|
|
|
|
a1b1 := fe_mul_generic(a1, b1)
|
|
a2b2 := fe_mul_generic(a2, b2)
|
|
assert a1b1 == a2b2 && is_in_bounds(a1b1) && is_in_bounds(a2b2)
|
|
}
|
|
}
|
|
|
|
fn test_fe_square_generic() {
|
|
for i in 0 .. 20 {
|
|
a := generate_field_element()
|
|
|
|
a1 := a
|
|
a2 := a
|
|
|
|
a11 := fe_square_generic(a1)
|
|
a22 := fe_square_generic(a2)
|
|
assert a11 == a22 && is_in_bounds(a11) && is_in_bounds(a22)
|
|
}
|
|
}
|
|
|
|
struct SqrtRatioTest {
|
|
u string
|
|
v string
|
|
was_square int
|
|
r string
|
|
}
|
|
|
|
fn test_sqrt_ratio() ? {
|
|
// From draft-irtf-cfrg-ristretto255-decaf448-00, Appendix A.4.
|
|
|
|
tests := [
|
|
// If u is 0, the function is defined to return (0, TRUE), even if v
|
|
// is zero. Note that where used in this package, the denominator v
|
|
// is never zero.
|
|
SqrtRatioTest{'0000000000000000000000000000000000000000000000000000000000000000', '0000000000000000000000000000000000000000000000000000000000000000', 1, '0000000000000000000000000000000000000000000000000000000000000000'},
|
|
// 0/1 == 0²
|
|
SqrtRatioTest{'0000000000000000000000000000000000000000000000000000000000000000', '0100000000000000000000000000000000000000000000000000000000000000', 1, '0000000000000000000000000000000000000000000000000000000000000000'},
|
|
// If u is non-zero and v is zero, defined to return (0, FALSE).
|
|
SqrtRatioTest{'0100000000000000000000000000000000000000000000000000000000000000', '0000000000000000000000000000000000000000000000000000000000000000', 0, '0000000000000000000000000000000000000000000000000000000000000000'},
|
|
// 2/1 is not square in this edwards25519.
|
|
SqrtRatioTest{'0200000000000000000000000000000000000000000000000000000000000000', '0100000000000000000000000000000000000000000000000000000000000000', 0, '3c5ff1b5d8e4113b871bd052f9e7bcd0582804c266ffb2d4f4203eb07fdb7c54'},
|
|
// 4/1 == 2²
|
|
SqrtRatioTest{'0400000000000000000000000000000000000000000000000000000000000000', '0100000000000000000000000000000000000000000000000000000000000000', 1, '0200000000000000000000000000000000000000000000000000000000000000'},
|
|
// 1/4 == (2⁻¹)² == (2^(p-2))² per Euler's theorem
|
|
SqrtRatioTest{'0100000000000000000000000000000000000000000000000000000000000000', '0400000000000000000000000000000000000000000000000000000000000000', 1, 'f6ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff3f'},
|
|
]
|
|
|
|
for i, tt in tests {
|
|
mut elu := Element{}
|
|
mut elv := Element{}
|
|
mut elw := Element{}
|
|
mut elg := Element{}
|
|
|
|
u := elu.set_bytes(hex.decode(tt.u) ?) ?
|
|
v := elv.set_bytes(hex.decode(tt.v) ?) ?
|
|
want := elw.set_bytes(hex.decode(tt.r) ?) ?
|
|
mut got, was_square := elg.sqrt_ratio(u, v)
|
|
|
|
assert got.equal(want) != 0
|
|
assert was_square == tt.was_square
|
|
// if got.Equal(want) == 0 || wasSquare != tt.wasSquare {
|
|
// t.Errorf("%d: got (%v, %v), want (%v, %v)", i, got, wasSquare, want, tt.wasSquare)
|
|
// }
|
|
}
|
|
}
|
|
|
|
fn test_set_bytes_normal() ? {
|
|
for i in 0 .. 15 {
|
|
mut el := Element{}
|
|
mut random_inp := rand.bytes(32) ?
|
|
|
|
el = el.set_bytes(random_inp.clone()) ?
|
|
random_inp[random_inp.len - 1] &= (1 << 7) - 1
|
|
// assert f1(random_inp, el) == true
|
|
|
|
assert random_inp == el.bytes()
|
|
assert is_in_bounds(el) == true
|
|
}
|
|
}
|
|
|
|
fn test_set_bytes_reduced() {
|
|
mut fe := Element{}
|
|
mut r := Element{}
|
|
mut random_inp := rand.bytes(32) or { return }
|
|
|
|
fe.set_bytes(random_inp) or { return }
|
|
r.set_bytes(fe.bytes()) or { return }
|
|
|
|
assert fe == r
|
|
}
|
|
|
|
// Check some fixed vectors from dalek
|
|
struct FeRTTest {
|
|
mut:
|
|
fe Element
|
|
b []byte
|
|
}
|
|
|
|
fn test_set_bytes_from_dalek_test_vectors() ? {
|
|
mut tests := [
|
|
FeRTTest{
|
|
fe: Element{358744748052810, 1691584618240980, 977650209285361, 1429865912637724, 560044844278676}
|
|
b: [byte(74), 209, 69, 197, 70, 70, 161, 222, 56, 226, 229, 19, 112, 60, 25, 92, 187,
|
|
74, 222, 56, 50, 153, 51, 233, 40, 74, 57, 6, 160, 185, 213, 31]
|
|
},
|
|
FeRTTest{
|
|
fe: Element{84926274344903, 473620666599931, 365590438845504, 1028470286882429, 2146499180330972}
|
|
b: [byte(199), 23, 106, 112, 61, 77, 216, 79, 186, 60, 11, 118, 13, 16, 103, 15, 42,
|
|
32, 83, 250, 44, 57, 204, 198, 78, 199, 253, 119, 146, 172, 3, 122]
|
|
},
|
|
]
|
|
for _, mut tt in tests {
|
|
b := tt.fe.bytes()
|
|
mut el := Element{}
|
|
mut fe := el.set_bytes(tt.b) ?
|
|
|
|
assert b == tt.b
|
|
assert fe.equal(tt.fe) == 1
|
|
}
|
|
}
|
|
|
|
fn test_equal() {
|
|
mut x := Element{1, 1, 1, 1, 1}
|
|
y := Element{5, 4, 3, 2, 1}
|
|
|
|
mut eq1 := x.equal(x)
|
|
assert eq1 == 1
|
|
|
|
eq1 = x.equal(y)
|
|
assert eq1 == 0
|
|
}
|
|
|
|
fn test_invert() ? {
|
|
mut x := Element{1, 1, 1, 1, 1}
|
|
mut one := Element{1, 0, 0, 0, 0}
|
|
mut xinv := Element{}
|
|
mut r := Element{}
|
|
|
|
xinv.invert(x)
|
|
r.multiply(x, xinv)
|
|
r.reduce()
|
|
|
|
assert one == r
|
|
bytes := rand.bytes(32) or { return err }
|
|
|
|
x.set_bytes(bytes) ?
|
|
|
|
xinv.invert(x)
|
|
r.multiply(x, xinv)
|
|
r.reduce()
|
|
|
|
assert one == r
|
|
|
|
zero := Element{}
|
|
x.set(zero)
|
|
|
|
xx := xinv.invert(x)
|
|
assert xx == xinv
|
|
assert xinv.equal(zero) == 1
|
|
// s := if num % 2 == 0 { 'even' } else { 'odd' }
|
|
}
|
|
|
|
fn test_mult_32() {
|
|
for j in 0 .. 10 {
|
|
mut x := Element{}
|
|
mut t1 := Element{}
|
|
y := u32(0)
|
|
for i := 0; i < 100; i++ {
|
|
t1.mult_32(x, y)
|
|
}
|
|
mut ty := Element{}
|
|
ty.l0 = u64(y)
|
|
mut t2 := Element{}
|
|
for i := 0; i < 100; i++ {
|
|
t2.multiply(x, ty)
|
|
}
|
|
assert t1.equal(t2) == 1 && is_in_bounds(t1) && is_in_bounds(t2)
|
|
}
|
|
}
|
|
|
|
fn test_selected_and_swap() {
|
|
a := Element{358744748052810, 1691584618240980, 977650209285361, 1429865912637724, 560044844278676}
|
|
b := Element{84926274344903, 473620666599931, 365590438845504, 1028470286882429, 2146499180330972}
|
|
|
|
mut c := Element{}
|
|
mut d := Element{}
|
|
|
|
c.selected(a, b, 1)
|
|
d.selected(a, b, 0)
|
|
|
|
assert c.equal(a) == 1
|
|
assert d.equal(b) == 1
|
|
|
|
c.swap(mut d, 0)
|
|
assert c.equal(a) == 1
|
|
assert d.equal(b) == 1
|
|
|
|
c.swap(mut d, 1)
|
|
assert c.equal(b) == 1
|
|
assert d.equal(a) == 1
|
|
}
|
|
|
|
// Tests self-consistency between multiply and Square.
|
|
fn test_consistency_between_mult_and_square() {
|
|
mut x := Element{1, 1, 1, 1, 1}
|
|
mut x2 := Element{}
|
|
mut x2sq := Element{}
|
|
|
|
x2.multiply(x, x)
|
|
x2sq.square(x)
|
|
|
|
assert x2 == x2sq
|
|
|
|
bytes := rand.bytes(32) or { return }
|
|
x.set_bytes(bytes) or { return }
|
|
x2.multiply(x, x)
|
|
x2sq.square(x)
|
|
|
|
assert x2 == x2sq
|
|
}
|
|
|
|
// to_big_integer returns v as a big.Integer.
|
|
fn (mut v Element) to_big_integer() big.Integer {
|
|
buf := v.bytes()
|
|
return big.integer_from_bytes(buf)
|
|
}
|
|
|
|
// from_big_integer sets v = n, and returns v. The bit length of n must not exceed 256.
|
|
fn (mut v Element) from_big_integer(n big.Integer) ?Element {
|
|
if n.binary_str().len > 32 * 8 {
|
|
return error('invalid edwards25519 element input size')
|
|
}
|
|
mut bytes, _ := n.bytes()
|
|
swap_endianness(mut bytes) // SHOULD I SWAP IT?
|
|
v.set_bytes(bytes) ?
|
|
|
|
return v
|
|
}
|
|
|
|
fn (mut v Element) from_decimal_string(s string) ?Element {
|
|
num := big.integer_from_string(s) ?
|
|
|
|
v = v.from_big_integer(num) ?
|
|
return v
|
|
}
|
|
|
|
fn test_bytes_big_equivalence() ? {
|
|
mut inp := rand.bytes(32) ?
|
|
el := Element{}
|
|
mut fe := el.generate_element()
|
|
mut fe1 := el.generate_element()
|
|
|
|
fe.set_bytes(inp) or { panic(err) }
|
|
inp[inp.len - 1] &= (1 << 7) - 1 // mask the most significant bit
|
|
|
|
mut b := big.integer_from_bytes(swap_endianness(mut inp)) // need swap_endianness
|
|
fe1.from_big_integer(b) or { panic(err) } // do swap_endianness internally
|
|
|
|
assert fe == fe1
|
|
|
|
mut buf := []byte{len: 32} // pad with zeroes
|
|
fedtobig := fe1.to_big_integer()
|
|
mut fedbig_bytes, _ := fedtobig.bytes()
|
|
copy(mut buf, fedbig_bytes) // does not need to do swap_endianness
|
|
|
|
assert fe.bytes() == buf && is_in_bounds(fe) && is_in_bounds(fe1)
|
|
// assert big_equivalence(inp, fe, fe1) == true
|
|
}
|
|
|
|
fn test_decimal_constants() ? {
|
|
sqrtm1string := '19681161376707505956807079304988542015446066515923890162744021073123829784752'
|
|
mut el := Element{}
|
|
mut exp := el.from_decimal_string(sqrtm1string) ?
|
|
|
|
assert sqrt_m1.equal(exp) == 1
|
|
|
|
dstring := '37095705934669439343138083508754565189542113879843219016388785533085940283555'
|
|
exp = el.from_decimal_string(dstring) ?
|
|
mut d := d_const
|
|
|
|
assert d.equal(exp) == 1
|
|
}
|
|
|
|
fn test_mul_64_to_128() {
|
|
mut a := u64(5)
|
|
mut b := u64(5)
|
|
mut r := mul_64(a, b)
|
|
|
|
assert r.lo == 0x19
|
|
assert r.hi == 0
|
|
|
|
a = u64(18014398509481983) // 2^54 - 1
|
|
b = u64(18014398509481983) // 2^54 - 1
|
|
r = mul_64(a, b)
|
|
|
|
assert r.lo == 0xff80000000000001 && r.hi == 0xfffffffffff
|
|
|
|
a = u64(1125899906842661)
|
|
b = u64(2097155)
|
|
r = mul_64(a, b)
|
|
r = add_mul_64(r, a, b)
|
|
r = add_mul_64(r, a, b)
|
|
r = add_mul_64(r, a, b)
|
|
r = add_mul_64(r, a, b)
|
|
|
|
assert r.lo == 16888498990613035 && r.hi == 640
|
|
}
|
|
|
|
fn test_multiply_distributes_over_add() {
|
|
for i in 0 .. 10 {
|
|
el := Element{}
|
|
x := el.generate_element()
|
|
y := el.generate_element()
|
|
z := el.generate_element()
|
|
mut t1 := Element{}
|
|
t1.add(x, y)
|
|
t1.multiply(t1, z)
|
|
|
|
// Compute t2 = x*z + y*z
|
|
mut t2 := Element{}
|
|
mut t3 := Element{}
|
|
t2.multiply(x, z)
|
|
t3.multiply(y, z)
|
|
t2.add(t2, t3)
|
|
assert t1.equal(t2) == 1 && is_in_bounds(t1) && is_in_bounds(t2)
|
|
}
|
|
}
|