Mirror
/
cpython
mirror of https://github.com/python/cpython
4
1
Fork 0
Code Issues Releases Wiki Activity
cpython/Modules/_hacl/Hacl_Hash_Blake2s.c

1515 lines
42 KiB
C
Raw Permalink Normal View History

gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
/* MIT License
*
* Copyright (c) 2016-2022 INRIA, CMU and Microsoft Corporation
* Copyright (c) 2022-2023 HACL* Contributors
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "internal/Hacl_Hash_Blake2s.h"
#include "internal/Hacl_Impl_Blake2_Constants.h"
#include "internal/Hacl_Hash_Blake2b.h"
#include "lib_memzero0.h"
static inline void
update_block(
uint32_t *wv,
uint32_t *hash,
bool flag,
bool last_node,
uint64_t totlen,
uint8_t *d
)
{
uint32_t m_w[16U] = { 0U };
KRML_MAYBE_FOR16(i,
0U,
16U,
1U,
uint32_t *os = m_w;
uint8_t *bj = d + i * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i] = x;);
uint32_t mask[4U] = { 0U };
uint32_t wv_14;
if (flag)
{
wv_14 = 0xFFFFFFFFU;
}
else
{
wv_14 = 0U;
}
uint32_t wv_15;
if (last_node)
{
wv_15 = 0xFFFFFFFFU;
}
else
{
wv_15 = 0U;
}
mask[0U] = (uint32_t)totlen;
mask[1U] = (uint32_t)(totlen >> 32U);
mask[2U] = wv_14;
mask[3U] = wv_15;
memcpy(wv, hash, 16U * sizeof (uint32_t));
uint32_t *wv3 = wv + 12U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv3;
uint32_t x = wv3[i] ^ mask[i];
os[i] = x;);
KRML_MAYBE_FOR10(i0,
0U,
10U,
1U,
uint32_t start_idx = i0 % 10U * 16U;
uint32_t m_st[16U] = { 0U };
uint32_t *r0 = m_st;
uint32_t *r1 = m_st + 4U;
uint32_t *r20 = m_st + 8U;
uint32_t *r30 = m_st + 12U;
uint32_t s0 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 0U];
uint32_t s1 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 1U];
uint32_t s2 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 2U];
uint32_t s3 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 3U];
uint32_t s4 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 4U];
uint32_t s5 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 5U];
uint32_t s6 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 6U];
uint32_t s7 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 7U];
uint32_t s8 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 8U];
uint32_t s9 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 9U];
uint32_t s10 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 10U];
uint32_t s11 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 11U];
uint32_t s12 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 12U];
uint32_t s13 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 13U];
uint32_t s14 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 14U];
uint32_t s15 = Hacl_Hash_Blake2b_sigmaTable[start_idx + 15U];
uint32_t uu____0 = m_w[s2];
uint32_t uu____1 = m_w[s4];
uint32_t uu____2 = m_w[s6];
r0[0U] = m_w[s0];
r0[1U] = uu____0;
r0[2U] = uu____1;
r0[3U] = uu____2;
uint32_t uu____3 = m_w[s3];
uint32_t uu____4 = m_w[s5];
uint32_t uu____5 = m_w[s7];
r1[0U] = m_w[s1];
r1[1U] = uu____3;
r1[2U] = uu____4;
r1[3U] = uu____5;
uint32_t uu____6 = m_w[s10];
uint32_t uu____7 = m_w[s12];
uint32_t uu____8 = m_w[s14];
r20[0U] = m_w[s8];
r20[1U] = uu____6;
r20[2U] = uu____7;
r20[3U] = uu____8;
uint32_t uu____9 = m_w[s11];
uint32_t uu____10 = m_w[s13];
uint32_t uu____11 = m_w[s15];
r30[0U] = m_w[s9];
r30[1U] = uu____9;
r30[2U] = uu____10;
r30[3U] = uu____11;
uint32_t *x = m_st;
uint32_t *y = m_st + 4U;
uint32_t *z = m_st + 8U;
uint32_t *w = m_st + 12U;
uint32_t a = 0U;
uint32_t b0 = 1U;
uint32_t c0 = 2U;
uint32_t d10 = 3U;
uint32_t *wv_a0 = wv + a * 4U;
uint32_t *wv_b0 = wv + b0 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a0;
uint32_t x1 = wv_a0[i] + wv_b0[i];
os[i] = x1;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a0;
uint32_t x1 = wv_a0[i] + x[i];
os[i] = x1;);
uint32_t *wv_a1 = wv + d10 * 4U;
uint32_t *wv_b1 = wv + a * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a1;
uint32_t x1 = wv_a1[i] ^ wv_b1[i];
os[i] = x1;);
uint32_t *r10 = wv_a1;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r10;
uint32_t x1 = r10[i];
uint32_t x10 = x1 >> 16U | x1 << 16U;
os[i] = x10;);
uint32_t *wv_a2 = wv + c0 * 4U;
uint32_t *wv_b2 = wv + d10 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a2;
uint32_t x1 = wv_a2[i] + wv_b2[i];
os[i] = x1;);
uint32_t *wv_a3 = wv + b0 * 4U;
uint32_t *wv_b3 = wv + c0 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a3;
uint32_t x1 = wv_a3[i] ^ wv_b3[i];
os[i] = x1;);
uint32_t *r12 = wv_a3;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r12;
uint32_t x1 = r12[i];
uint32_t x10 = x1 >> 12U | x1 << 20U;
os[i] = x10;);
uint32_t *wv_a4 = wv + a * 4U;
uint32_t *wv_b4 = wv + b0 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a4;
uint32_t x1 = wv_a4[i] + wv_b4[i];
os[i] = x1;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a4;
uint32_t x1 = wv_a4[i] + y[i];
os[i] = x1;);
uint32_t *wv_a5 = wv + d10 * 4U;
uint32_t *wv_b5 = wv + a * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a5;
uint32_t x1 = wv_a5[i] ^ wv_b5[i];
os[i] = x1;);
uint32_t *r13 = wv_a5;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r13;
uint32_t x1 = r13[i];
uint32_t x10 = x1 >> 8U | x1 << 24U;
os[i] = x10;);
uint32_t *wv_a6 = wv + c0 * 4U;
uint32_t *wv_b6 = wv + d10 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a6;
uint32_t x1 = wv_a6[i] + wv_b6[i];
os[i] = x1;);
uint32_t *wv_a7 = wv + b0 * 4U;
uint32_t *wv_b7 = wv + c0 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a7;
uint32_t x1 = wv_a7[i] ^ wv_b7[i];
os[i] = x1;);
uint32_t *r14 = wv_a7;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r14;
uint32_t x1 = r14[i];
uint32_t x10 = x1 >> 7U | x1 << 25U;
os[i] = x10;);
uint32_t *r15 = wv + 4U;
uint32_t *r21 = wv + 8U;
uint32_t *r31 = wv + 12U;
uint32_t *r110 = r15;
uint32_t x00 = r110[1U];
uint32_t x10 = r110[2U];
uint32_t x20 = r110[3U];
uint32_t x30 = r110[0U];
r110[0U] = x00;
r110[1U] = x10;
r110[2U] = x20;
r110[3U] = x30;
uint32_t *r111 = r21;
uint32_t x01 = r111[2U];
uint32_t x11 = r111[3U];
uint32_t x21 = r111[0U];
uint32_t x31 = r111[1U];
r111[0U] = x01;
r111[1U] = x11;
r111[2U] = x21;
r111[3U] = x31;
uint32_t *r112 = r31;
uint32_t x02 = r112[3U];
uint32_t x12 = r112[0U];
uint32_t x22 = r112[1U];
uint32_t x32 = r112[2U];
r112[0U] = x02;
r112[1U] = x12;
r112[2U] = x22;
r112[3U] = x32;
uint32_t a0 = 0U;
uint32_t b = 1U;
uint32_t c = 2U;
uint32_t d1 = 3U;
uint32_t *wv_a = wv + a0 * 4U;
uint32_t *wv_b8 = wv + b * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a;
uint32_t x1 = wv_a[i] + wv_b8[i];
os[i] = x1;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a;
uint32_t x1 = wv_a[i] + z[i];
os[i] = x1;);
uint32_t *wv_a8 = wv + d1 * 4U;
uint32_t *wv_b9 = wv + a0 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a8;
uint32_t x1 = wv_a8[i] ^ wv_b9[i];
os[i] = x1;);
uint32_t *r16 = wv_a8;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r16;
uint32_t x1 = r16[i];
uint32_t x13 = x1 >> 16U | x1 << 16U;
os[i] = x13;);
uint32_t *wv_a9 = wv + c * 4U;
uint32_t *wv_b10 = wv + d1 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a9;
uint32_t x1 = wv_a9[i] + wv_b10[i];
os[i] = x1;);
uint32_t *wv_a10 = wv + b * 4U;
uint32_t *wv_b11 = wv + c * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a10;
uint32_t x1 = wv_a10[i] ^ wv_b11[i];
os[i] = x1;);
uint32_t *r17 = wv_a10;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r17;
uint32_t x1 = r17[i];
uint32_t x13 = x1 >> 12U | x1 << 20U;
os[i] = x13;);
uint32_t *wv_a11 = wv + a0 * 4U;
uint32_t *wv_b12 = wv + b * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a11;
uint32_t x1 = wv_a11[i] + wv_b12[i];
os[i] = x1;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a11;
uint32_t x1 = wv_a11[i] + w[i];
os[i] = x1;);
uint32_t *wv_a12 = wv + d1 * 4U;
uint32_t *wv_b13 = wv + a0 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a12;
uint32_t x1 = wv_a12[i] ^ wv_b13[i];
os[i] = x1;);
uint32_t *r18 = wv_a12;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r18;
uint32_t x1 = r18[i];
uint32_t x13 = x1 >> 8U | x1 << 24U;
os[i] = x13;);
uint32_t *wv_a13 = wv + c * 4U;
uint32_t *wv_b14 = wv + d1 * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a13;
uint32_t x1 = wv_a13[i] + wv_b14[i];
os[i] = x1;);
uint32_t *wv_a14 = wv + b * 4U;
uint32_t *wv_b = wv + c * 4U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = wv_a14;
uint32_t x1 = wv_a14[i] ^ wv_b[i];
os[i] = x1;);
uint32_t *r19 = wv_a14;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = r19;
uint32_t x1 = r19[i];
uint32_t x13 = x1 >> 7U | x1 << 25U;
os[i] = x13;);
uint32_t *r113 = wv + 4U;
uint32_t *r2 = wv + 8U;
uint32_t *r3 = wv + 12U;
uint32_t *r11 = r113;
uint32_t x03 = r11[3U];
uint32_t x13 = r11[0U];
uint32_t x23 = r11[1U];
uint32_t x33 = r11[2U];
r11[0U] = x03;
r11[1U] = x13;
r11[2U] = x23;
r11[3U] = x33;
uint32_t *r114 = r2;
uint32_t x04 = r114[2U];
uint32_t x14 = r114[3U];
uint32_t x24 = r114[0U];
uint32_t x34 = r114[1U];
r114[0U] = x04;
r114[1U] = x14;
r114[2U] = x24;
r114[3U] = x34;
uint32_t *r115 = r3;
uint32_t x0 = r115[1U];
uint32_t x1 = r115[2U];
uint32_t x2 = r115[3U];
uint32_t x3 = r115[0U];
r115[0U] = x0;
r115[1U] = x1;
r115[2U] = x2;
r115[3U] = x3;);
uint32_t *s0 = hash;
uint32_t *s1 = hash + 4U;
uint32_t *r0 = wv;
uint32_t *r1 = wv + 4U;
uint32_t *r2 = wv + 8U;
uint32_t *r3 = wv + 12U;
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = s0;
uint32_t x = s0[i] ^ r0[i];
os[i] = x;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = s0;
uint32_t x = s0[i] ^ r2[i];
os[i] = x;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = s1;
uint32_t x = s1[i] ^ r1[i];
os[i] = x;);
KRML_MAYBE_FOR4(i,
0U,
4U,
1U,
uint32_t *os = s1;
uint32_t x = s1[i] ^ r3[i];
os[i] = x;);
}
void Hacl_Hash_Blake2s_init(uint32_t *hash, uint32_t kk, uint32_t nn)
{
uint8_t salt[8U] = { 0U };
uint8_t personal[8U] = { 0U };
Hacl_Hash_Blake2b_blake2_params
p =
{
.digest_length = 32U, .key_length = 0U, .fanout = 1U, .depth = 1U, .leaf_length = 0U,
.node_offset = 0ULL, .node_depth = 0U, .inner_length = 0U, .salt = salt, .personal = personal
};
uint32_t tmp[8U] = { 0U };
uint32_t *r0 = hash;
uint32_t *r1 = hash + 4U;
uint32_t *r2 = hash + 8U;
uint32_t *r3 = hash + 12U;
uint32_t iv0 = Hacl_Hash_Blake2b_ivTable_S[0U];
uint32_t iv1 = Hacl_Hash_Blake2b_ivTable_S[1U];
uint32_t iv2 = Hacl_Hash_Blake2b_ivTable_S[2U];
uint32_t iv3 = Hacl_Hash_Blake2b_ivTable_S[3U];
uint32_t iv4 = Hacl_Hash_Blake2b_ivTable_S[4U];
uint32_t iv5 = Hacl_Hash_Blake2b_ivTable_S[5U];
uint32_t iv6 = Hacl_Hash_Blake2b_ivTable_S[6U];
uint32_t iv7 = Hacl_Hash_Blake2b_ivTable_S[7U];
r2[0U] = iv0;
r2[1U] = iv1;
r2[2U] = iv2;
r2[3U] = iv3;
r3[0U] = iv4;
r3[1U] = iv5;
r3[2U] = iv6;
r3[3U] = iv7;
KRML_MAYBE_FOR2(i,
0U,
2U,
1U,
uint32_t *os = tmp + 4U;
uint8_t *bj = p.salt + i * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i] = x;);
KRML_MAYBE_FOR2(i,
0U,
2U,
1U,
uint32_t *os = tmp + 6U;
uint8_t *bj = p.personal + i * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i] = x;);
tmp[0U] =
(uint32_t)(uint8_t)nn
^ ((uint32_t)(uint8_t)kk << 8U ^ ((uint32_t)p.fanout << 16U ^ (uint32_t)p.depth << 24U));
tmp[1U] = p.leaf_length;
tmp[2U] = (uint32_t)p.node_offset;
tmp[3U] =
(uint32_t)(p.node_offset >> 32U)
^ ((uint32_t)p.node_depth << 16U ^ (uint32_t)p.inner_length << 24U);
uint32_t tmp0 = tmp[0U];
uint32_t tmp1 = tmp[1U];
uint32_t tmp2 = tmp[2U];
uint32_t tmp3 = tmp[3U];
uint32_t tmp4 = tmp[4U];
uint32_t tmp5 = tmp[5U];
uint32_t tmp6 = tmp[6U];
uint32_t tmp7 = tmp[7U];
uint32_t iv0_ = iv0 ^ tmp0;
uint32_t iv1_ = iv1 ^ tmp1;
uint32_t iv2_ = iv2 ^ tmp2;
uint32_t iv3_ = iv3 ^ tmp3;
uint32_t iv4_ = iv4 ^ tmp4;
uint32_t iv5_ = iv5 ^ tmp5;
uint32_t iv6_ = iv6 ^ tmp6;
uint32_t iv7_ = iv7 ^ tmp7;
r0[0U] = iv0_;
r0[1U] = iv1_;
r0[2U] = iv2_;
r0[3U] = iv3_;
r1[0U] = iv4_;
r1[1U] = iv5_;
r1[2U] = iv6_;
r1[3U] = iv7_;
}
static void update_key(uint32_t *wv, uint32_t *hash, uint32_t kk, uint8_t *k, uint32_t ll)
{
uint64_t lb = (uint64_t)64U;
uint8_t b[64U] = { 0U };
memcpy(b, k, kk * sizeof (uint8_t));
if (ll == 0U)
{
update_block(wv, hash, true, false, lb, b);
}
else
{
update_block(wv, hash, false, false, lb, b);
}
Lib_Memzero0_memzero(b, 64U, uint8_t, void *);
}
void
Hacl_Hash_Blake2s_update_multi(
uint32_t len,
uint32_t *wv,
uint32_t *hash,
uint64_t prev,
uint8_t *blocks,
uint32_t nb
)
{
KRML_MAYBE_UNUSED_VAR(len);
for (uint32_t i = 0U; i < nb; i++)
{
uint64_t totlen = prev + (uint64_t)((i + 1U) * 64U);
uint8_t *b = blocks + i * 64U;
update_block(wv, hash, false, false, totlen, b);
}
}
void
Hacl_Hash_Blake2s_update_last(
uint32_t len,
uint32_t *wv,
uint32_t *hash,
bool last_node,
uint64_t prev,
uint32_t rem,
uint8_t *d
)
{
uint8_t b[64U] = { 0U };
uint8_t *last = d + len - rem;
memcpy(b, last, rem * sizeof (uint8_t));
uint64_t totlen = prev + (uint64_t)len;
update_block(wv, hash, true, last_node, totlen, b);
Lib_Memzero0_memzero(b, 64U, uint8_t, void *);
}
static void
update_blocks(uint32_t len, uint32_t *wv, uint32_t *hash, uint64_t prev, uint8_t *blocks)
{
uint32_t nb0 = len / 64U;
uint32_t rem0 = len % 64U;
uint32_t nb;
if (rem0 == 0U && nb0 > 0U)
{
nb = nb0 - 1U;
}
else
{
nb = nb0;
}
uint32_t rem;
if (rem0 == 0U && nb0 > 0U)
{
rem = 64U;
}
else
{
rem = rem0;
}
Hacl_Hash_Blake2s_update_multi(len, wv, hash, prev, blocks, nb);
Hacl_Hash_Blake2s_update_last(len, wv, hash, false, prev, rem, blocks);
}
static inline void
update(uint32_t *wv, uint32_t *hash, uint32_t kk, uint8_t *k, uint32_t ll, uint8_t *d)
{
uint64_t lb = (uint64_t)64U;
if (kk > 0U)
{
update_key(wv, hash, kk, k, ll);
if (!(ll == 0U))
{
update_blocks(ll, wv, hash, lb, d);
return;
}
return;
}
update_blocks(ll, wv, hash, (uint64_t)0U, d);
}
void Hacl_Hash_Blake2s_finish(uint32_t nn, uint8_t *output, uint32_t *hash)
{
uint8_t b[32U] = { 0U };
uint8_t *first = b;
uint8_t *second = b + 16U;
uint32_t *row0 = hash;
uint32_t *row1 = hash + 4U;
KRML_MAYBE_FOR4(i, 0U, 4U, 1U, store32_le(first + i * 4U, row0[i]););
KRML_MAYBE_FOR4(i, 0U, 4U, 1U, store32_le(second + i * 4U, row1[i]););
uint8_t *final = b;
memcpy(output, final, nn * sizeof (uint8_t));
Lib_Memzero0_memzero(b, 32U, uint8_t, void *);
}
static Hacl_Hash_Blake2s_state_t
*malloc_raw(Hacl_Hash_Blake2b_index kk, Hacl_Hash_Blake2b_params_and_key key)
{
uint8_t *buf = (uint8_t *)KRML_HOST_CALLOC(64U, sizeof (uint8_t));
uint32_t *wv = (uint32_t *)KRML_HOST_CALLOC(16U, sizeof (uint32_t));
uint32_t *b = (uint32_t *)KRML_HOST_CALLOC(16U, sizeof (uint32_t));
Hacl_Hash_Blake2s_block_state_t
block_state =
{
.fst = kk.key_length,
.snd = kk.digest_length,
.thd = kk.last_node,
.f3 = { .fst = wv, .snd = b }
};
uint8_t kk10 = kk.key_length;
uint32_t ite;
if (kk10 != 0U)
{
ite = 64U;
}
else
{
ite = 0U;
}
Hacl_Hash_Blake2s_state_t
s = { .block_state = block_state, .buf = buf, .total_len = (uint64_t)ite };
Hacl_Hash_Blake2s_state_t
*p = (Hacl_Hash_Blake2s_state_t *)KRML_HOST_MALLOC(sizeof (Hacl_Hash_Blake2s_state_t));
p[0U] = s;
Hacl_Hash_Blake2b_blake2_params *p1 = key.fst;
uint8_t kk1 = p1->key_length;
uint8_t nn = p1->digest_length;
bool last_node = block_state.thd;
Hacl_Hash_Blake2b_index i = { .key_length = kk1, .digest_length = nn, .last_node = last_node };
GH-99108: Make vectorized versions of Blake2 available on x86, too (#125244) Accomplished by updating HACL* vendored code from hacl-star/hacl-star@a6a09496d9cff652b567d26f2c3ab012321b632a to hacl-star/hacl-star@315a9e491d2bc347b9dae99e0ea506995ea84d9d Co-authored-by: Victor Stinner <vstinner@python.org> Co-authored-by: Zachary Ware <zach@python.org>
2024-10-17 12:08:43 -03:00
uint32_t *h = block_state.f3.snd;
gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
uint32_t kk2 = (uint32_t)i.key_length;
uint8_t *k_1 = key.snd;
if (!(kk2 == 0U))
{
uint8_t *sub_b = buf + kk2;
memset(sub_b, 0U, (64U - kk2) * sizeof (uint8_t));
memcpy(buf, k_1, kk2 * sizeof (uint8_t));
}
Hacl_Hash_Blake2b_blake2_params pv = p1[0U];
GH-99108: Make vectorized versions of Blake2 available on x86, too (#125244) Accomplished by updating HACL* vendored code from hacl-star/hacl-star@a6a09496d9cff652b567d26f2c3ab012321b632a to hacl-star/hacl-star@315a9e491d2bc347b9dae99e0ea506995ea84d9d Co-authored-by: Victor Stinner <vstinner@python.org> Co-authored-by: Zachary Ware <zach@python.org>
2024-10-17 12:08:43 -03:00
uint32_t tmp[8U] = { 0U };
uint32_t *r0 = h;
uint32_t *r1 = h + 4U;
uint32_t *r2 = h + 8U;
uint32_t *r3 = h + 12U;
uint32_t iv0 = Hacl_Hash_Blake2b_ivTable_S[0U];
uint32_t iv1 = Hacl_Hash_Blake2b_ivTable_S[1U];
uint32_t iv2 = Hacl_Hash_Blake2b_ivTable_S[2U];
uint32_t iv3 = Hacl_Hash_Blake2b_ivTable_S[3U];
uint32_t iv4 = Hacl_Hash_Blake2b_ivTable_S[4U];
uint32_t iv5 = Hacl_Hash_Blake2b_ivTable_S[5U];
uint32_t iv6 = Hacl_Hash_Blake2b_ivTable_S[6U];
uint32_t iv7 = Hacl_Hash_Blake2b_ivTable_S[7U];
r2[0U] = iv0;
r2[1U] = iv1;
r2[2U] = iv2;
r2[3U] = iv3;
r3[0U] = iv4;
r3[1U] = iv5;
r3[2U] = iv6;
r3[3U] = iv7;
KRML_MAYBE_FOR2(i0,
0U,
2U,
1U,
uint32_t *os = tmp + 4U;
uint8_t *bj = pv.salt + i0 * 4U;
uint32_t u = load32_le(bj);
uint32_t r4 = u;
uint32_t x = r4;
os[i0] = x;);
KRML_MAYBE_FOR2(i0,
0U,
2U,
1U,
uint32_t *os = tmp + 6U;
uint8_t *bj = pv.personal + i0 * 4U;
uint32_t u = load32_le(bj);
uint32_t r4 = u;
uint32_t x = r4;
os[i0] = x;);
tmp[0U] =
(uint32_t)pv.digest_length
^ ((uint32_t)pv.key_length << 8U ^ ((uint32_t)pv.fanout << 16U ^ (uint32_t)pv.depth << 24U));
tmp[1U] = pv.leaf_length;
tmp[2U] = (uint32_t)pv.node_offset;
tmp[3U] =
(uint32_t)(pv.node_offset >> 32U)
^ ((uint32_t)pv.node_depth << 16U ^ (uint32_t)pv.inner_length << 24U);
uint32_t tmp0 = tmp[0U];
uint32_t tmp1 = tmp[1U];
uint32_t tmp2 = tmp[2U];
uint32_t tmp3 = tmp[3U];
uint32_t tmp4 = tmp[4U];
uint32_t tmp5 = tmp[5U];
uint32_t tmp6 = tmp[6U];
uint32_t tmp7 = tmp[7U];
uint32_t iv0_ = iv0 ^ tmp0;
uint32_t iv1_ = iv1 ^ tmp1;
uint32_t iv2_ = iv2 ^ tmp2;
uint32_t iv3_ = iv3 ^ tmp3;
uint32_t iv4_ = iv4 ^ tmp4;
uint32_t iv5_ = iv5 ^ tmp5;
uint32_t iv6_ = iv6 ^ tmp6;
uint32_t iv7_ = iv7 ^ tmp7;
r0[0U] = iv0_;
r0[1U] = iv1_;
r0[2U] = iv2_;
r0[3U] = iv3_;
r1[0U] = iv4_;
r1[1U] = iv5_;
r1[2U] = iv6_;
r1[3U] = iv7_;
gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
return p;
}
/**
General-purpose allocation function that gives control over all
Blake2 parameters, including the key. Further resettings of the state SHALL be
done with `reset_with_params_and_key`, and SHALL feature the exact same values
for the `key_length` and `digest_length` fields as passed here. In other words,
once you commit to a digest and key length, the only way to change these
parameters is to allocate a new object.
The caller must satisfy the following requirements.
- The length of the key k MUST match the value of the field key_length in the
parameters.
- The key_length must not exceed 32 for S, 64 for B.
- The digest_length must not exceed 32 for S, 64 for B.
*/
Hacl_Hash_Blake2s_state_t
*Hacl_Hash_Blake2s_malloc_with_params_and_key(
Hacl_Hash_Blake2b_blake2_params *p,
bool last_node,
uint8_t *k
)
{
Hacl_Hash_Blake2b_blake2_params pv = p[0U];
Hacl_Hash_Blake2b_index
i1 = { .key_length = pv.key_length, .digest_length = pv.digest_length, .last_node = last_node };
return malloc_raw(i1, ((Hacl_Hash_Blake2b_params_and_key){ .fst = p, .snd = k }));
}
/**
Specialized allocation function that picks default values for all
parameters, except for the key_length. Further resettings of the state SHALL be
done with `reset_with_key`, and SHALL feature the exact same key length `kk` as
passed here. In other words, once you commit to a key length, the only way to
change this parameter is to allocate a new object.
The caller must satisfy the following requirements.
- The key_length must not exceed 32 for S, 64 for B.
*/
Hacl_Hash_Blake2s_state_t *Hacl_Hash_Blake2s_malloc_with_key(uint8_t *k, uint8_t kk)
{
uint8_t nn = 32U;
Hacl_Hash_Blake2b_index i = { .key_length = kk, .digest_length = nn, .last_node = false };
uint8_t salt[8U] = { 0U };
uint8_t personal[8U] = { 0U };
Hacl_Hash_Blake2b_blake2_params
p =
{
.digest_length = i.digest_length, .key_length = i.key_length, .fanout = 1U, .depth = 1U,
.leaf_length = 0U, .node_offset = 0ULL, .node_depth = 0U, .inner_length = 0U, .salt = salt,
.personal = personal
};
Hacl_Hash_Blake2b_blake2_params p0 = p;
Hacl_Hash_Blake2s_state_t *s = Hacl_Hash_Blake2s_malloc_with_params_and_key(&p0, false, k);
return s;
}
/**
Specialized allocation function that picks default values for all
parameters, and has no key. Effectively, this is what you want if you intend to
use Blake2 as a hash function. Further resettings of the state SHALL be done with `reset`.
*/
Hacl_Hash_Blake2s_state_t *Hacl_Hash_Blake2s_malloc(void)
{
return Hacl_Hash_Blake2s_malloc_with_key(NULL, 0U);
}
static Hacl_Hash_Blake2b_index index_of_state(Hacl_Hash_Blake2s_state_t *s)
{
Hacl_Hash_Blake2s_block_state_t block_state = (*s).block_state;
bool last_node = block_state.thd;
uint8_t nn = block_state.snd;
uint8_t kk1 = block_state.fst;
return
((Hacl_Hash_Blake2b_index){ .key_length = kk1, .digest_length = nn, .last_node = last_node });
}
static void reset_raw(Hacl_Hash_Blake2s_state_t *state, Hacl_Hash_Blake2b_params_and_key key)
{
Hacl_Hash_Blake2s_state_t scrut = *state;
uint8_t *buf = scrut.buf;
Hacl_Hash_Blake2s_block_state_t block_state = scrut.block_state;
bool last_node0 = block_state.thd;
uint8_t nn0 = block_state.snd;
uint8_t kk10 = block_state.fst;
Hacl_Hash_Blake2b_index
i = { .key_length = kk10, .digest_length = nn0, .last_node = last_node0 };
KRML_MAYBE_UNUSED_VAR(i);
Hacl_Hash_Blake2b_blake2_params *p = key.fst;
uint8_t kk1 = p->key_length;
uint8_t nn = p->digest_length;
bool last_node = block_state.thd;
Hacl_Hash_Blake2b_index
i1 = { .key_length = kk1, .digest_length = nn, .last_node = last_node };
GH-99108: Make vectorized versions of Blake2 available on x86, too (#125244) Accomplished by updating HACL* vendored code from hacl-star/hacl-star@a6a09496d9cff652b567d26f2c3ab012321b632a to hacl-star/hacl-star@315a9e491d2bc347b9dae99e0ea506995ea84d9d Co-authored-by: Victor Stinner <vstinner@python.org> Co-authored-by: Zachary Ware <zach@python.org>
2024-10-17 12:08:43 -03:00
uint32_t *h = block_state.f3.snd;
gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
uint32_t kk2 = (uint32_t)i1.key_length;
uint8_t *k_1 = key.snd;
if (!(kk2 == 0U))
{
uint8_t *sub_b = buf + kk2;
memset(sub_b, 0U, (64U - kk2) * sizeof (uint8_t));
memcpy(buf, k_1, kk2 * sizeof (uint8_t));
}
Hacl_Hash_Blake2b_blake2_params pv = p[0U];
GH-99108: Make vectorized versions of Blake2 available on x86, too (#125244) Accomplished by updating HACL* vendored code from hacl-star/hacl-star@a6a09496d9cff652b567d26f2c3ab012321b632a to hacl-star/hacl-star@315a9e491d2bc347b9dae99e0ea506995ea84d9d Co-authored-by: Victor Stinner <vstinner@python.org> Co-authored-by: Zachary Ware <zach@python.org>
2024-10-17 12:08:43 -03:00
uint32_t tmp[8U] = { 0U };
uint32_t *r0 = h;
uint32_t *r1 = h + 4U;
uint32_t *r2 = h + 8U;
uint32_t *r3 = h + 12U;
uint32_t iv0 = Hacl_Hash_Blake2b_ivTable_S[0U];
uint32_t iv1 = Hacl_Hash_Blake2b_ivTable_S[1U];
uint32_t iv2 = Hacl_Hash_Blake2b_ivTable_S[2U];
uint32_t iv3 = Hacl_Hash_Blake2b_ivTable_S[3U];
uint32_t iv4 = Hacl_Hash_Blake2b_ivTable_S[4U];
uint32_t iv5 = Hacl_Hash_Blake2b_ivTable_S[5U];
uint32_t iv6 = Hacl_Hash_Blake2b_ivTable_S[6U];
uint32_t iv7 = Hacl_Hash_Blake2b_ivTable_S[7U];
r2[0U] = iv0;
r2[1U] = iv1;
r2[2U] = iv2;
r2[3U] = iv3;
r3[0U] = iv4;
r3[1U] = iv5;
r3[2U] = iv6;
r3[3U] = iv7;
KRML_MAYBE_FOR2(i0,
0U,
2U,
1U,
uint32_t *os = tmp + 4U;
uint8_t *bj = pv.salt + i0 * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i0] = x;);
KRML_MAYBE_FOR2(i0,
0U,
2U,
1U,
uint32_t *os = tmp + 6U;
uint8_t *bj = pv.personal + i0 * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i0] = x;);
tmp[0U] =
(uint32_t)pv.digest_length
^ ((uint32_t)pv.key_length << 8U ^ ((uint32_t)pv.fanout << 16U ^ (uint32_t)pv.depth << 24U));
tmp[1U] = pv.leaf_length;
tmp[2U] = (uint32_t)pv.node_offset;
tmp[3U] =
(uint32_t)(pv.node_offset >> 32U)
^ ((uint32_t)pv.node_depth << 16U ^ (uint32_t)pv.inner_length << 24U);
uint32_t tmp0 = tmp[0U];
uint32_t tmp1 = tmp[1U];
uint32_t tmp2 = tmp[2U];
uint32_t tmp3 = tmp[3U];
uint32_t tmp4 = tmp[4U];
uint32_t tmp5 = tmp[5U];
uint32_t tmp6 = tmp[6U];
uint32_t tmp7 = tmp[7U];
uint32_t iv0_ = iv0 ^ tmp0;
uint32_t iv1_ = iv1 ^ tmp1;
uint32_t iv2_ = iv2 ^ tmp2;
uint32_t iv3_ = iv3 ^ tmp3;
uint32_t iv4_ = iv4 ^ tmp4;
uint32_t iv5_ = iv5 ^ tmp5;
uint32_t iv6_ = iv6 ^ tmp6;
uint32_t iv7_ = iv7 ^ tmp7;
r0[0U] = iv0_;
r0[1U] = iv1_;
r0[2U] = iv2_;
r0[3U] = iv3_;
r1[0U] = iv4_;
r1[1U] = iv5_;
r1[2U] = iv6_;
r1[3U] = iv7_;
gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
uint8_t kk11 = i.key_length;
uint32_t ite;
if (kk11 != 0U)
{
ite = 64U;
}
else
{
ite = 0U;
}
Hacl_Hash_Blake2s_state_t
GH-99108: Make vectorized versions of Blake2 available on x86, too (#125244) Accomplished by updating HACL* vendored code from hacl-star/hacl-star@a6a09496d9cff652b567d26f2c3ab012321b632a to hacl-star/hacl-star@315a9e491d2bc347b9dae99e0ea506995ea84d9d Co-authored-by: Victor Stinner <vstinner@python.org> Co-authored-by: Zachary Ware <zach@python.org>
2024-10-17 12:08:43 -03:00
tmp8 = { .block_state = block_state, .buf = buf, .total_len = (uint64_t)ite };
state[0U] = tmp8;
gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
}
/**
General-purpose re-initialization function with parameters and
key. You cannot change digest_length, key_length, or last_node, meaning those values in
the parameters object must be the same as originally decided via one of the
malloc functions. All other values of the parameter can be changed. The behavior
is unspecified if you violate this precondition.
*/
void
Hacl_Hash_Blake2s_reset_with_key_and_params(
Hacl_Hash_Blake2s_state_t *s,
Hacl_Hash_Blake2b_blake2_params *p,
uint8_t *k
)
{
gh-126433: Update hacl-star (GH-126791) Retrieve the change: "Lib_Memzero0.c: Fix compiler warning on 32-bit Windows".
2024-11-14 20:22:50 -04:00
Hacl_Hash_Blake2b_index i1 = index_of_state(s);
KRML_MAYBE_UNUSED_VAR(i1);
gh-99108: Add HACL* Blake2 implementation to hashlib (GH-119316) This replaces the existing hashlib Blake2 module with a single implementation that uses HACL\*'s Blake2b/Blake2s implementations. We added support for all the modes exposed by the Python API, including tree hashing, leaf nodes, and so on. We ported and merged all of these changes upstream in HACL\*, added test vectors based on Python's existing implementation, and exposed everything needed for hashlib. This was joint work done with @R1kM. See the PR for much discussion and benchmarking details. TL;DR: On many systems, 8-50% faster (!) than `libb2`, on some systems it appeared 10-20% slower than `libb2`.
2024-08-13 18:42:19 -03:00
reset_raw(s, ((Hacl_Hash_Blake2b_params_and_key){ .fst = p, .snd = k }));
}
/**
Specialized-purpose re-initialization function with no parameters,
and a key. The key length must be the same as originally decided via your choice
of malloc function. All other parameters are reset to their default values. The
original call to malloc MUST have set digest_length to the default value. The
behavior is unspecified if you violate this precondition.
*/
void Hacl_Hash_Blake2s_reset_with_key(Hacl_Hash_Blake2s_state_t *s, uint8_t *k)
{
Hacl_Hash_Blake2b_index idx = index_of_state(s);
uint8_t salt[8U] = { 0U };
uint8_t personal[8U] = { 0U };
Hacl_Hash_Blake2b_blake2_params
p =
{
.digest_length = idx.digest_length, .key_length = idx.key_length, .fanout = 1U, .depth = 1U,
.leaf_length = 0U, .node_offset = 0ULL, .node_depth = 0U, .inner_length = 0U, .salt = salt,
.personal = personal
};
Hacl_Hash_Blake2b_blake2_params p0 = p;
reset_raw(s, ((Hacl_Hash_Blake2b_params_and_key){ .fst = &p0, .snd = k }));
}
/**
Specialized-purpose re-initialization function with no parameters
and no key. This is what you want if you intend to use Blake2 as a hash
function. The key length and digest length must have been set to their
respective default values via your choice of malloc function (always true if you
used `malloc`). All other parameters are reset to their default values. The
behavior is unspecified if you violate this precondition.
*/
void Hacl_Hash_Blake2s_reset(Hacl_Hash_Blake2s_state_t *s)
{
Hacl_Hash_Blake2s_reset_with_key(s, NULL);
}
/**
Update function; 0 = success, 1 = max length exceeded
*/
Hacl_Streaming_Types_error_code
Hacl_Hash_Blake2s_update(Hacl_Hash_Blake2s_state_t *state, uint8_t *chunk, uint32_t chunk_len)
{
Hacl_Hash_Blake2s_state_t s = *state;
uint64_t total_len = s.total_len;
if ((uint64_t)chunk_len > 0xffffffffffffffffULL - total_len)
{
return Hacl_Streaming_Types_MaximumLengthExceeded;
}
uint32_t sz;
if (total_len % (uint64_t)64U == 0ULL && total_len > 0ULL)
{
sz = 64U;
}
else
{
sz = (uint32_t)(total_len % (uint64_t)64U);
}
if (chunk_len <= 64U - sz)
{
Hacl_Hash_Blake2s_state_t s1 = *state;
Hacl_Hash_Blake2s_block_state_t block_state1 = s1.block_state;
uint8_t *buf = s1.buf;
uint64_t total_len1 = s1.total_len;
uint32_t sz1;
if (total_len1 % (uint64_t)64U == 0ULL && total_len1 > 0ULL)
{
sz1 = 64U;
}
else
{
sz1 = (uint32_t)(total_len1 % (uint64_t)64U);
}
uint8_t *buf2 = buf + sz1;
memcpy(buf2, chunk, chunk_len * sizeof (uint8_t));
uint64_t total_len2 = total_len1 + (uint64_t)chunk_len;
*state
=
(
(Hacl_Hash_Blake2s_state_t){
.block_state = block_state1,
.buf = buf,
.total_len = total_len2
}
);
}
else if (sz == 0U)
{
Hacl_Hash_Blake2s_state_t s1 = *state;
Hacl_Hash_Blake2s_block_state_t block_state1 = s1.block_state;
uint8_t *buf = s1.buf;
uint64_t total_len1 = s1.total_len;
uint32_t sz1;
if (total_len1 % (uint64_t)64U == 0ULL && total_len1 > 0ULL)
{
sz1 = 64U;
}
else
{
sz1 = (uint32_t)(total_len1 % (uint64_t)64U);
}
if (!(sz1 == 0U))
{
uint64_t prevlen = total_len1 - (uint64_t)sz1;
K____uint32_t___uint32_t_ acc = block_state1.f3;
uint32_t *wv = acc.fst;
uint32_t *hash = acc.snd;
uint32_t nb = 1U;
Hacl_Hash_Blake2s_update_multi(64U, wv, hash, prevlen, buf, nb);
}
uint32_t ite;
if ((uint64_t)chunk_len % (uint64_t)64U == 0ULL && (uint64_t)chunk_len > 0ULL)
{
ite = 64U;
}
else
{
ite = (uint32_t)((uint64_t)chunk_len % (uint64_t)64U);
}
uint32_t n_blocks = (chunk_len - ite) / 64U;
uint32_t data1_len = n_blocks * 64U;
uint32_t data2_len = chunk_len - data1_len;
uint8_t *data1 = chunk;
uint8_t *data2 = chunk + data1_len;
K____uint32_t___uint32_t_ acc = block_state1.f3;
uint32_t *wv = acc.fst;
uint32_t *hash = acc.snd;
uint32_t nb = data1_len / 64U;
Hacl_Hash_Blake2s_update_multi(data1_len, wv, hash, total_len1, data1, nb);
uint8_t *dst = buf;
memcpy(dst, data2, data2_len * sizeof (uint8_t));
*state
=
(
(Hacl_Hash_Blake2s_state_t){
.block_state = block_state1,
.buf = buf,
.total_len = total_len1 + (uint64_t)chunk_len
}
);
}
else
{
uint32_t diff = 64U - sz;
uint8_t *chunk1 = chunk;
uint8_t *chunk2 = chunk + diff;
Hacl_Hash_Blake2s_state_t s1 = *state;
Hacl_Hash_Blake2s_block_state_t block_state10 = s1.block_state;
uint8_t *buf0 = s1.buf;
uint64_t total_len10 = s1.total_len;
uint32_t sz10;
if (total_len10 % (uint64_t)64U == 0ULL && total_len10 > 0ULL)
{
sz10 = 64U;
}
else
{
sz10 = (uint32_t)(total_len10 % (uint64_t)64U);
}
uint8_t *buf2 = buf0 + sz10;
memcpy(buf2, chunk1, diff * sizeof (uint8_t));
uint64_t total_len2 = total_len10 + (uint64_t)diff;
*state
=
(
(Hacl_Hash_Blake2s_state_t){
.block_state = block_state10,
.buf = buf0,
.total_len = total_len2
}
);
Hacl_Hash_Blake2s_state_t s10 = *state;
Hacl_Hash_Blake2s_block_state_t block_state1 = s10.block_state;
uint8_t *buf = s10.buf;
uint64_t total_len1 = s10.total_len;
uint32_t sz1;
if (total_len1 % (uint64_t)64U == 0ULL && total_len1 > 0ULL)
{
sz1 = 64U;
}
else
{
sz1 = (uint32_t)(total_len1 % (uint64_t)64U);
}
if (!(sz1 == 0U))
{
uint64_t prevlen = total_len1 - (uint64_t)sz1;
K____uint32_t___uint32_t_ acc = block_state1.f3;
uint32_t *wv = acc.fst;
uint32_t *hash = acc.snd;
uint32_t nb = 1U;
Hacl_Hash_Blake2s_update_multi(64U, wv, hash, prevlen, buf, nb);
}
uint32_t ite;
if
((uint64_t)(chunk_len - diff) % (uint64_t)64U == 0ULL && (uint64_t)(chunk_len - diff) > 0ULL)
{
ite = 64U;
}
else
{
ite = (uint32_t)((uint64_t)(chunk_len - diff) % (uint64_t)64U);
}
uint32_t n_blocks = (chunk_len - diff - ite) / 64U;
uint32_t data1_len = n_blocks * 64U;
uint32_t data2_len = chunk_len - diff - data1_len;
uint8_t *data1 = chunk2;
uint8_t *data2 = chunk2 + data1_len;
K____uint32_t___uint32_t_ acc = block_state1.f3;
uint32_t *wv = acc.fst;
uint32_t *hash = acc.snd;
uint32_t nb = data1_len / 64U;
Hacl_Hash_Blake2s_update_multi(data1_len, wv, hash, total_len1, data1, nb);
uint8_t *dst = buf;
memcpy(dst, data2, data2_len * sizeof (uint8_t));
*state
=
(
(Hacl_Hash_Blake2s_state_t){
.block_state = block_state1,
.buf = buf,
.total_len = total_len1 + (uint64_t)(chunk_len - diff)
}
);
}
return Hacl_Streaming_Types_Success;
}
/**
Digest function. This function expects the `output` array to hold
at least `digest_length` bytes, where `digest_length` was determined by your
choice of `malloc` function. Concretely, if you used `malloc` or
`malloc_with_key`, then the expected length is 32 for S, or 64 for B (default
digest length). If you used `malloc_with_params_and_key`, then the expected
length is whatever you chose for the `digest_length` field of your parameters.
For convenience, this function returns `digest_length`. When in doubt, callers
can pass an array of size HACL_BLAKE2S_32_OUT_BYTES, then use the return value
to see how many bytes were actually written.
*/
uint8_t Hacl_Hash_Blake2s_digest(Hacl_Hash_Blake2s_state_t *s, uint8_t *dst)
{
Hacl_Hash_Blake2s_block_state_t block_state0 = (*s).block_state;
bool last_node0 = block_state0.thd;
uint8_t nn0 = block_state0.snd;
uint8_t kk0 = block_state0.fst;
Hacl_Hash_Blake2b_index
i1 = { .key_length = kk0, .digest_length = nn0, .last_node = last_node0 };
Hacl_Hash_Blake2s_state_t scrut = *s;
Hacl_Hash_Blake2s_block_state_t block_state = scrut.block_state;
uint8_t *buf_ = scrut.buf;
uint64_t total_len = scrut.total_len;
uint32_t r;
if (total_len % (uint64_t)64U == 0ULL && total_len > 0ULL)
{
r = 64U;
}
else
{
r = (uint32_t)(total_len % (uint64_t)64U);
}
uint8_t *buf_1 = buf_;
uint32_t wv0[16U] = { 0U };
uint32_t b[16U] = { 0U };
Hacl_Hash_Blake2s_block_state_t
tmp_block_state =
{
.fst = i1.key_length,
.snd = i1.digest_length,
.thd = i1.last_node,
.f3 = { .fst = wv0, .snd = b }
};
uint32_t *src_b = block_state.f3.snd;
uint32_t *dst_b = tmp_block_state.f3.snd;
memcpy(dst_b, src_b, 16U * sizeof (uint32_t));
uint64_t prev_len = total_len - (uint64_t)r;
uint32_t ite;
if (r % 64U == 0U && r > 0U)
{
ite = 64U;
}
else
{
ite = r % 64U;
}
uint8_t *buf_last = buf_1 + r - ite;
uint8_t *buf_multi = buf_1;
K____uint32_t___uint32_t_ acc0 = tmp_block_state.f3;
uint32_t *wv1 = acc0.fst;
uint32_t *hash0 = acc0.snd;
uint32_t nb = 0U;
Hacl_Hash_Blake2s_update_multi(0U, wv1, hash0, prev_len, buf_multi, nb);
uint64_t prev_len_last = total_len - (uint64_t)r;
K____uint32_t___uint32_t_ acc = tmp_block_state.f3;
bool last_node1 = tmp_block_state.thd;
uint32_t *wv = acc.fst;
uint32_t *hash = acc.snd;
Hacl_Hash_Blake2s_update_last(r, wv, hash, last_node1, prev_len_last, r, buf_last);
uint8_t nn1 = tmp_block_state.snd;
Hacl_Hash_Blake2s_finish((uint32_t)nn1, dst, tmp_block_state.f3.snd);
Hacl_Hash_Blake2s_block_state_t block_state1 = (*s).block_state;
bool last_node = block_state1.thd;
uint8_t nn = block_state1.snd;
uint8_t kk = block_state1.fst;
return
((Hacl_Hash_Blake2b_index){ .key_length = kk, .digest_length = nn, .last_node = last_node }).digest_length;
}
Hacl_Hash_Blake2b_index Hacl_Hash_Blake2s_info(Hacl_Hash_Blake2s_state_t *s)
{
Hacl_Hash_Blake2s_block_state_t block_state = (*s).block_state;
bool last_node = block_state.thd;
uint8_t nn = block_state.snd;
uint8_t kk = block_state.fst;
return
((Hacl_Hash_Blake2b_index){ .key_length = kk, .digest_length = nn, .last_node = last_node });
}
/**
Free state function when there is no key
*/
void Hacl_Hash_Blake2s_free(Hacl_Hash_Blake2s_state_t *state)
{
Hacl_Hash_Blake2s_state_t scrut = *state;
uint8_t *buf = scrut.buf;
Hacl_Hash_Blake2s_block_state_t block_state = scrut.block_state;
uint32_t *b = block_state.f3.snd;
uint32_t *wv = block_state.f3.fst;
KRML_HOST_FREE(wv);
KRML_HOST_FREE(b);
KRML_HOST_FREE(buf);
KRML_HOST_FREE(state);
}
/**
Copying. This preserves all parameters.
*/
Hacl_Hash_Blake2s_state_t *Hacl_Hash_Blake2s_copy(Hacl_Hash_Blake2s_state_t *state)
{
Hacl_Hash_Blake2s_state_t scrut = *state;
Hacl_Hash_Blake2s_block_state_t block_state0 = scrut.block_state;
uint8_t *buf0 = scrut.buf;
uint64_t total_len0 = scrut.total_len;
bool last_node = block_state0.thd;
uint8_t nn = block_state0.snd;
uint8_t kk1 = block_state0.fst;
Hacl_Hash_Blake2b_index i = { .key_length = kk1, .digest_length = nn, .last_node = last_node };
uint8_t *buf = (uint8_t *)KRML_HOST_CALLOC(64U, sizeof (uint8_t));
memcpy(buf, buf0, 64U * sizeof (uint8_t));
uint32_t *wv = (uint32_t *)KRML_HOST_CALLOC(16U, sizeof (uint32_t));
uint32_t *b = (uint32_t *)KRML_HOST_CALLOC(16U, sizeof (uint32_t));
Hacl_Hash_Blake2s_block_state_t
block_state =
{
.fst = i.key_length,
.snd = i.digest_length,
.thd = i.last_node,
.f3 = { .fst = wv, .snd = b }
};
uint32_t *src_b = block_state0.f3.snd;
uint32_t *dst_b = block_state.f3.snd;
memcpy(dst_b, src_b, 16U * sizeof (uint32_t));
Hacl_Hash_Blake2s_state_t
s = { .block_state = block_state, .buf = buf, .total_len = total_len0 };
Hacl_Hash_Blake2s_state_t
*p = (Hacl_Hash_Blake2s_state_t *)KRML_HOST_MALLOC(sizeof (Hacl_Hash_Blake2s_state_t));
p[0U] = s;
return p;
}
/**
Write the BLAKE2s digest of message `input` using key `key` into `output`.
@param output Pointer to `output_len` bytes of memory where the digest is written to.
@param output_len Length of the to-be-generated digest with 1 <= `output_len` <= 64.
@param input Pointer to `input_len` bytes of memory where the input message is read from.
@param input_len Length of the input message.
@param key Pointer to `key_len` bytes of memory where the key is read from.
@param key_len Length of the key. Can be 0.
*/
void
Hacl_Hash_Blake2s_hash_with_key(
uint8_t *output,
uint32_t output_len,
uint8_t *input,
uint32_t input_len,
uint8_t *key,
uint32_t key_len
)
{
uint32_t b[16U] = { 0U };
uint32_t b1[16U] = { 0U };
Hacl_Hash_Blake2s_init(b, key_len, output_len);
update(b1, b, key_len, key, input_len, input);
Hacl_Hash_Blake2s_finish(output_len, output, b);
Lib_Memzero0_memzero(b1, 16U, uint32_t, void *);
Lib_Memzero0_memzero(b, 16U, uint32_t, void *);
}
/**
Write the BLAKE2s digest of message `input` using key `key` and
parameters `params` into `output`. The `key` array must be of length
`params.key_length`. The `output` array must be of length
`params.digest_length`.
*/
void
Hacl_Hash_Blake2s_hash_with_key_and_params(
uint8_t *output,
uint8_t *input,
uint32_t input_len,
Hacl_Hash_Blake2b_blake2_params params,
uint8_t *key
)
{
uint32_t b[16U] = { 0U };
uint32_t b1[16U] = { 0U };
uint32_t tmp[8U] = { 0U };
uint32_t *r0 = b;
uint32_t *r1 = b + 4U;
uint32_t *r2 = b + 8U;
uint32_t *r3 = b + 12U;
uint32_t iv0 = Hacl_Hash_Blake2b_ivTable_S[0U];
uint32_t iv1 = Hacl_Hash_Blake2b_ivTable_S[1U];
uint32_t iv2 = Hacl_Hash_Blake2b_ivTable_S[2U];
uint32_t iv3 = Hacl_Hash_Blake2b_ivTable_S[3U];
uint32_t iv4 = Hacl_Hash_Blake2b_ivTable_S[4U];
uint32_t iv5 = Hacl_Hash_Blake2b_ivTable_S[5U];
uint32_t iv6 = Hacl_Hash_Blake2b_ivTable_S[6U];
uint32_t iv7 = Hacl_Hash_Blake2b_ivTable_S[7U];
r2[0U] = iv0;
r2[1U] = iv1;
r2[2U] = iv2;
r2[3U] = iv3;
r3[0U] = iv4;
r3[1U] = iv5;
r3[2U] = iv6;
r3[3U] = iv7;
KRML_MAYBE_FOR2(i,
0U,
2U,
1U,
uint32_t *os = tmp + 4U;
uint8_t *bj = params.salt + i * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i] = x;);
KRML_MAYBE_FOR2(i,
0U,
2U,
1U,
uint32_t *os = tmp + 6U;
uint8_t *bj = params.personal + i * 4U;
uint32_t u = load32_le(bj);
uint32_t r = u;
uint32_t x = r;
os[i] = x;);
tmp[0U] =
(uint32_t)params.digest_length
^
((uint32_t)params.key_length
<< 8U
^ ((uint32_t)params.fanout << 16U ^ (uint32_t)params.depth << 24U));
tmp[1U] = params.leaf_length;
tmp[2U] = (uint32_t)params.node_offset;
tmp[3U] =
(uint32_t)(params.node_offset >> 32U)
^ ((uint32_t)params.node_depth << 16U ^ (uint32_t)params.inner_length << 24U);
uint32_t tmp0 = tmp[0U];
uint32_t tmp1 = tmp[1U];
uint32_t tmp2 = tmp[2U];
uint32_t tmp3 = tmp[3U];
uint32_t tmp4 = tmp[4U];
uint32_t tmp5 = tmp[5U];
uint32_t tmp6 = tmp[6U];
uint32_t tmp7 = tmp[7U];
uint32_t iv0_ = iv0 ^ tmp0;
uint32_t iv1_ = iv1 ^ tmp1;
uint32_t iv2_ = iv2 ^ tmp2;
uint32_t iv3_ = iv3 ^ tmp3;
uint32_t iv4_ = iv4 ^ tmp4;
uint32_t iv5_ = iv5 ^ tmp5;
uint32_t iv6_ = iv6 ^ tmp6;
uint32_t iv7_ = iv7 ^ tmp7;
r0[0U] = iv0_;
r0[1U] = iv1_;
r0[2U] = iv2_;
r0[3U] = iv3_;
r1[0U] = iv4_;
r1[1U] = iv5_;
r1[2U] = iv6_;
r1[3U] = iv7_;
update(b1, b, (uint32_t)params.key_length, key, input_len, input);
Hacl_Hash_Blake2s_finish((uint32_t)params.digest_length, output, b);
Lib_Memzero0_memzero(b1, 16U, uint32_t, void *);
Lib_Memzero0_memzero(b, 16U, uint32_t, void *);
}