lib/crypto/poly1305-donna32.c

   1 // SPDX-License-Identifier: GPL-2.0 OR MIT
   2 /*
   3  * Copyright (C) 2015-2019 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
   4  *
   5  * This is based in part on Andrew Moon's poly1305-donna, which is in the
   6  * public domain.
   7  */
   8
   9 #include <linux/kernel.h>
  10 #include <asm/unaligned.h>
  11 #include <crypto/internal/poly1305.h>
  12
  13 void poly1305_core_setkey(struct poly1305_core_key *key, const u8 raw_key[16])
  14 {
  15         /* r &= 0xffffffc0ffffffc0ffffffc0fffffff */
  16         key->key.r[0] = (get_unaligned_le32(&raw_key[0])) & 0x3ffffff;
  17         key->key.r[1] = (get_unaligned_le32(&raw_key[3]) >> 2) & 0x3ffff03;
  18         key->key.r[2] = (get_unaligned_le32(&raw_key[6]) >> 4) & 0x3ffc0ff;
  19         key->key.r[3] = (get_unaligned_le32(&raw_key[9]) >> 6) & 0x3f03fff;
  20         key->key.r[4] = (get_unaligned_le32(&raw_key[12]) >> 8) & 0x00fffff;
  21
  22         /* s = 5*r */
  23         key->precomputed_s.r[0] = key->key.r[1] * 5;
  24         key->precomputed_s.r[1] = key->key.r[2] * 5;
  25         key->precomputed_s.r[2] = key->key.r[3] * 5;
  26         key->precomputed_s.r[3] = key->key.r[4] * 5;
  27 }
  28 EXPORT_SYMBOL(poly1305_core_setkey);
  29
  30 void poly1305_core_blocks(struct poly1305_state *state,
  31                           const struct poly1305_core_key *key, const void *src,
  32                           unsigned int nblocks, u32 hibit)
  33 {
  34         const u8 *input = src;
  35         u32 r0, r1, r2, r3, r4;
  36         u32 s1, s2, s3, s4;
  37         u32 h0, h1, h2, h3, h4;
  38         u64 d0, d1, d2, d3, d4;
  39         u32 c;
  40
  41         if (!nblocks)
  42                 return;
  43
  44         hibit <<= 24;
  45
  46         r0 = key->key.r[0];
  47         r1 = key->key.r[1];
  48         r2 = key->key.r[2];
  49         r3 = key->key.r[3];
  50         r4 = key->key.r[4];
  51
  52         s1 = key->precomputed_s.r[0];
  53         s2 = key->precomputed_s.r[1];
  54         s3 = key->precomputed_s.r[2];
  55         s4 = key->precomputed_s.r[3];
  56
  57         h0 = state->h[0];
  58         h1 = state->h[1];
  59         h2 = state->h[2];
  60         h3 = state->h[3];
  61         h4 = state->h[4];
  62
  63         do {
  64                 /* h += m[i] */
  65                 h0 += (get_unaligned_le32(&input[0])) & 0x3ffffff;
  66                 h1 += (get_unaligned_le32(&input[3]) >> 2) & 0x3ffffff;
  67                 h2 += (get_unaligned_le32(&input[6]) >> 4) & 0x3ffffff;
  68                 h3 += (get_unaligned_le32(&input[9]) >> 6) & 0x3ffffff;
  69                 h4 += (get_unaligned_le32(&input[12]) >> 8) | hibit;
  70
  71                 /* h *= r */
  72                 d0 = ((u64)h0 * r0) + ((u64)h1 * s4) +
  73                      ((u64)h2 * s3) + ((u64)h3 * s2) +
  74                      ((u64)h4 * s1);
  75                 d1 = ((u64)h0 * r1) + ((u64)h1 * r0) +
  76                      ((u64)h2 * s4) + ((u64)h3 * s3) +
  77                      ((u64)h4 * s2);
  78                 d2 = ((u64)h0 * r2) + ((u64)h1 * r1) +
  79                      ((u64)h2 * r0) + ((u64)h3 * s4) +
  80                      ((u64)h4 * s3);
  81                 d3 = ((u64)h0 * r3) + ((u64)h1 * r2) +
  82                      ((u64)h2 * r1) + ((u64)h3 * r0) +
  83                      ((u64)h4 * s4);
  84                 d4 = ((u64)h0 * r4) + ((u64)h1 * r3) +
  85                      ((u64)h2 * r2) + ((u64)h3 * r1) +
  86                      ((u64)h4 * r0);
  87
  88                 /* (partial) h %= p */
  89                 c = (u32)(d0 >> 26);
  90                 h0 = (u32)d0 & 0x3ffffff;
  91                 d1 += c;
  92                 c = (u32)(d1 >> 26);
  93                 h1 = (u32)d1 & 0x3ffffff;
  94                 d2 += c;
  95                 c = (u32)(d2 >> 26);
  96                 h2 = (u32)d2 & 0x3ffffff;
  97                 d3 += c;
  98                 c = (u32)(d3 >> 26);
  99                 h3 = (u32)d3 & 0x3ffffff;
 100                 d4 += c;
 101                 c = (u32)(d4 >> 26);
 102                 h4 = (u32)d4 & 0x3ffffff;
 103                 h0 += c * 5;
 104                 c = (h0 >> 26);
 105                 h0 = h0 & 0x3ffffff;
 106                 h1 += c;
 107
 108                 input += POLY1305_BLOCK_SIZE;
 109         } while (--nblocks);
 110
 111         state->h[0] = h0;
 112         state->h[1] = h1;
 113         state->h[2] = h2;
 114         state->h[3] = h3;
 115         state->h[4] = h4;
 116 }
 117 EXPORT_SYMBOL(poly1305_core_blocks);
 118
 119 void poly1305_core_emit(const struct poly1305_state *state, const u32 nonce[4],
 120                         void *dst)
 121 {
 122         u8 *mac = dst;
 123         u32 h0, h1, h2, h3, h4, c;
 124         u32 g0, g1, g2, g3, g4;
 125         u64 f;
 126         u32 mask;
 127
 128         /* fully carry h */
 129         h0 = state->h[0];
 130         h1 = state->h[1];
 131         h2 = state->h[2];
 132         h3 = state->h[3];
 133         h4 = state->h[4];
 134
 135         c = h1 >> 26;
 136         h1 = h1 & 0x3ffffff;
 137         h2 += c;
 138         c = h2 >> 26;
 139         h2 = h2 & 0x3ffffff;
 140         h3 += c;
 141         c = h3 >> 26;
 142         h3 = h3 & 0x3ffffff;
 143         h4 += c;
 144         c = h4 >> 26;
 145         h4 = h4 & 0x3ffffff;
 146         h0 += c * 5;
 147         c = h0 >> 26;
 148         h0 = h0 & 0x3ffffff;
 149         h1 += c;
 150
 151         /* compute h + -p */
 152         g0 = h0 + 5;
 153         c = g0 >> 26;
 154         g0 &= 0x3ffffff;
 155         g1 = h1 + c;
 156         c = g1 >> 26;
 157         g1 &= 0x3ffffff;
 158         g2 = h2 + c;
 159         c = g2 >> 26;
 160         g2 &= 0x3ffffff;
 161         g3 = h3 + c;
 162         c = g3 >> 26;
 163         g3 &= 0x3ffffff;
 164         g4 = h4 + c - (1UL << 26);
 165
 166         /* select h if h < p, or h + -p if h >= p */
 167         mask = (g4 >> ((sizeof(u32) * 8) - 1)) - 1;
 168         g0 &= mask;
 169         g1 &= mask;
 170         g2 &= mask;
 171         g3 &= mask;
 172         g4 &= mask;
 173         mask = ~mask;
 174
 175         h0 = (h0 & mask) | g0;
 176         h1 = (h1 & mask) | g1;
 177         h2 = (h2 & mask) | g2;
 178         h3 = (h3 & mask) | g3;
 179         h4 = (h4 & mask) | g4;
 180
 181         /* h = h % (2^128) */
 182         h0 = ((h0) | (h1 << 26)) & 0xffffffff;
 183         h1 = ((h1 >> 6) | (h2 << 20)) & 0xffffffff;
 184         h2 = ((h2 >> 12) | (h3 << 14)) & 0xffffffff;
 185         h3 = ((h3 >> 18) | (h4 << 8)) & 0xffffffff;
 186
 187         if (likely(nonce)) {
 188                 /* mac = (h + nonce) % (2^128) */
 189                 f = (u64)h0 + nonce[0];
 190                 h0 = (u32)f;
 191                 f = (u64)h1 + nonce[1] + (f >> 32);
 192                 h1 = (u32)f;
 193                 f = (u64)h2 + nonce[2] + (f >> 32);
 194                 h2 = (u32)f;
 195                 f = (u64)h3 + nonce[3] + (f >> 32);
 196                 h3 = (u32)f;
 197         }
 198
 199         put_unaligned_le32(h0, &mac[0]);
 200         put_unaligned_le32(h1, &mac[4]);
 201         put_unaligned_le32(h2, &mac[8]);
 202         put_unaligned_le32(h3, &mac[12]);
 203 }
 204 EXPORT_SYMBOL(poly1305_core_emit);