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