1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Poly1305 authenticator algorithm, RFC7539 4 * 5 * Copyright (C) 2015 Martin Willi 6 * 7 * Based on public domain code by Andrew Moon and Daniel J. Bernstein. 8 */ 9 10 #include <crypto/internal/poly1305.h> 11 #include <linux/kernel.h> 12 #include <linux/module.h> 13 #include <asm/unaligned.h> 14 15 static inline u64 mlt(u64 a, u64 b) 16 { 17 return a * b; 18 } 19 20 static inline u32 sr(u64 v, u_char n) 21 { 22 return v >> n; 23 } 24 25 static inline u32 and(u32 v, u32 mask) 26 { 27 return v & mask; 28 } 29 30 void poly1305_core_setkey(struct poly1305_key *key, const u8 *raw_key) 31 { 32 /* r &= 0xffffffc0ffffffc0ffffffc0fffffff */ 33 key->r[0] = (get_unaligned_le32(raw_key + 0) >> 0) & 0x3ffffff; 34 key->r[1] = (get_unaligned_le32(raw_key + 3) >> 2) & 0x3ffff03; 35 key->r[2] = (get_unaligned_le32(raw_key + 6) >> 4) & 0x3ffc0ff; 36 key->r[3] = (get_unaligned_le32(raw_key + 9) >> 6) & 0x3f03fff; 37 key->r[4] = (get_unaligned_le32(raw_key + 12) >> 8) & 0x00fffff; 38 } 39 EXPORT_SYMBOL_GPL(poly1305_core_setkey); 40 41 void poly1305_core_blocks(struct poly1305_state *state, 42 const struct poly1305_key *key, const void *src, 43 unsigned int nblocks, u32 hibit) 44 { 45 u32 r0, r1, r2, r3, r4; 46 u32 s1, s2, s3, s4; 47 u32 h0, h1, h2, h3, h4; 48 u64 d0, d1, d2, d3, d4; 49 50 if (!nblocks) 51 return; 52 53 r0 = key->r[0]; 54 r1 = key->r[1]; 55 r2 = key->r[2]; 56 r3 = key->r[3]; 57 r4 = key->r[4]; 58 59 s1 = r1 * 5; 60 s2 = r2 * 5; 61 s3 = r3 * 5; 62 s4 = r4 * 5; 63 64 h0 = state->h[0]; 65 h1 = state->h[1]; 66 h2 = state->h[2]; 67 h3 = state->h[3]; 68 h4 = state->h[4]; 69 70 do { 71 /* h += m[i] */ 72 h0 += (get_unaligned_le32(src + 0) >> 0) & 0x3ffffff; 73 h1 += (get_unaligned_le32(src + 3) >> 2) & 0x3ffffff; 74 h2 += (get_unaligned_le32(src + 6) >> 4) & 0x3ffffff; 75 h3 += (get_unaligned_le32(src + 9) >> 6) & 0x3ffffff; 76 h4 += (get_unaligned_le32(src + 12) >> 8) | (hibit << 24); 77 78 /* h *= r */ 79 d0 = mlt(h0, r0) + mlt(h1, s4) + mlt(h2, s3) + 80 mlt(h3, s2) + mlt(h4, s1); 81 d1 = mlt(h0, r1) + mlt(h1, r0) + mlt(h2, s4) + 82 mlt(h3, s3) + mlt(h4, s2); 83 d2 = mlt(h0, r2) + mlt(h1, r1) + mlt(h2, r0) + 84 mlt(h3, s4) + mlt(h4, s3); 85 d3 = mlt(h0, r3) + mlt(h1, r2) + mlt(h2, r1) + 86 mlt(h3, r0) + mlt(h4, s4); 87 d4 = mlt(h0, r4) + mlt(h1, r3) + mlt(h2, r2) + 88 mlt(h3, r1) + mlt(h4, r0); 89 90 /* (partial) h %= p */ 91 d1 += sr(d0, 26); h0 = and(d0, 0x3ffffff); 92 d2 += sr(d1, 26); h1 = and(d1, 0x3ffffff); 93 d3 += sr(d2, 26); h2 = and(d2, 0x3ffffff); 94 d4 += sr(d3, 26); h3 = and(d3, 0x3ffffff); 95 h0 += sr(d4, 26) * 5; h4 = and(d4, 0x3ffffff); 96 h1 += h0 >> 26; h0 = h0 & 0x3ffffff; 97 98 src += POLY1305_BLOCK_SIZE; 99 } while (--nblocks); 100 101 state->h[0] = h0; 102 state->h[1] = h1; 103 state->h[2] = h2; 104 state->h[3] = h3; 105 state->h[4] = h4; 106 } 107 EXPORT_SYMBOL_GPL(poly1305_core_blocks); 108 109 void poly1305_core_emit(const struct poly1305_state *state, void *dst) 110 { 111 u32 h0, h1, h2, h3, h4; 112 u32 g0, g1, g2, g3, g4; 113 u32 mask; 114 115 /* fully carry h */ 116 h0 = state->h[0]; 117 h1 = state->h[1]; 118 h2 = state->h[2]; 119 h3 = state->h[3]; 120 h4 = state->h[4]; 121 122 h2 += (h1 >> 26); h1 = h1 & 0x3ffffff; 123 h3 += (h2 >> 26); h2 = h2 & 0x3ffffff; 124 h4 += (h3 >> 26); h3 = h3 & 0x3ffffff; 125 h0 += (h4 >> 26) * 5; h4 = h4 & 0x3ffffff; 126 h1 += (h0 >> 26); h0 = h0 & 0x3ffffff; 127 128 /* compute h + -p */ 129 g0 = h0 + 5; 130 g1 = h1 + (g0 >> 26); g0 &= 0x3ffffff; 131 g2 = h2 + (g1 >> 26); g1 &= 0x3ffffff; 132 g3 = h3 + (g2 >> 26); g2 &= 0x3ffffff; 133 g4 = h4 + (g3 >> 26) - (1 << 26); g3 &= 0x3ffffff; 134 135 /* select h if h < p, or h + -p if h >= p */ 136 mask = (g4 >> ((sizeof(u32) * 8) - 1)) - 1; 137 g0 &= mask; 138 g1 &= mask; 139 g2 &= mask; 140 g3 &= mask; 141 g4 &= mask; 142 mask = ~mask; 143 h0 = (h0 & mask) | g0; 144 h1 = (h1 & mask) | g1; 145 h2 = (h2 & mask) | g2; 146 h3 = (h3 & mask) | g3; 147 h4 = (h4 & mask) | g4; 148 149 /* h = h % (2^128) */ 150 put_unaligned_le32((h0 >> 0) | (h1 << 26), dst + 0); 151 put_unaligned_le32((h1 >> 6) | (h2 << 20), dst + 4); 152 put_unaligned_le32((h2 >> 12) | (h3 << 14), dst + 8); 153 put_unaligned_le32((h3 >> 18) | (h4 << 8), dst + 12); 154 } 155 EXPORT_SYMBOL_GPL(poly1305_core_emit); 156 157 void poly1305_init_generic(struct poly1305_desc_ctx *desc, const u8 *key) 158 { 159 poly1305_core_setkey(desc->r, key); 160 desc->s[0] = get_unaligned_le32(key + 16); 161 desc->s[1] = get_unaligned_le32(key + 20); 162 desc->s[2] = get_unaligned_le32(key + 24); 163 desc->s[3] = get_unaligned_le32(key + 28); 164 poly1305_core_init(&desc->h); 165 desc->buflen = 0; 166 desc->sset = true; 167 desc->rset = 1; 168 } 169 EXPORT_SYMBOL_GPL(poly1305_init_generic); 170 171 void poly1305_update_generic(struct poly1305_desc_ctx *desc, const u8 *src, 172 unsigned int nbytes) 173 { 174 unsigned int bytes; 175 176 if (unlikely(desc->buflen)) { 177 bytes = min(nbytes, POLY1305_BLOCK_SIZE - desc->buflen); 178 memcpy(desc->buf + desc->buflen, src, bytes); 179 src += bytes; 180 nbytes -= bytes; 181 desc->buflen += bytes; 182 183 if (desc->buflen == POLY1305_BLOCK_SIZE) { 184 poly1305_core_blocks(&desc->h, desc->r, desc->buf, 1, 1); 185 desc->buflen = 0; 186 } 187 } 188 189 if (likely(nbytes >= POLY1305_BLOCK_SIZE)) { 190 poly1305_core_blocks(&desc->h, desc->r, src, 191 nbytes / POLY1305_BLOCK_SIZE, 1); 192 src += nbytes - (nbytes % POLY1305_BLOCK_SIZE); 193 nbytes %= POLY1305_BLOCK_SIZE; 194 } 195 196 if (unlikely(nbytes)) { 197 desc->buflen = nbytes; 198 memcpy(desc->buf, src, nbytes); 199 } 200 } 201 EXPORT_SYMBOL_GPL(poly1305_update_generic); 202 203 void poly1305_final_generic(struct poly1305_desc_ctx *desc, u8 *dst) 204 { 205 __le32 digest[4]; 206 u64 f = 0; 207 208 if (unlikely(desc->buflen)) { 209 desc->buf[desc->buflen++] = 1; 210 memset(desc->buf + desc->buflen, 0, 211 POLY1305_BLOCK_SIZE - desc->buflen); 212 poly1305_core_blocks(&desc->h, desc->r, desc->buf, 1, 0); 213 } 214 215 poly1305_core_emit(&desc->h, digest); 216 217 /* mac = (h + s) % (2^128) */ 218 f = (f >> 32) + le32_to_cpu(digest[0]) + desc->s[0]; 219 put_unaligned_le32(f, dst + 0); 220 f = (f >> 32) + le32_to_cpu(digest[1]) + desc->s[1]; 221 put_unaligned_le32(f, dst + 4); 222 f = (f >> 32) + le32_to_cpu(digest[2]) + desc->s[2]; 223 put_unaligned_le32(f, dst + 8); 224 f = (f >> 32) + le32_to_cpu(digest[3]) + desc->s[3]; 225 put_unaligned_le32(f, dst + 12); 226 227 *desc = (struct poly1305_desc_ctx){}; 228 } 229 EXPORT_SYMBOL_GPL(poly1305_final_generic); 230 231 MODULE_LICENSE("GPL"); 232 MODULE_AUTHOR("Martin Willi <martin@strongswan.org>"); 233