diff options
Diffstat (limited to 'test/monniaux/BearSSL/src/hash')
-rw-r--r-- | test/monniaux/BearSSL/src/hash/dig_oid.c | 84 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/dig_size.c | 50 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/ghash_ctmul.c | 345 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/ghash_ctmul32.c | 251 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/ghash_ctmul64.c | 154 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/ghash_pclmul.c | 389 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/ghash_pwr8.c | 411 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/md5.c | 208 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/md5sha1.c | 141 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/mgf1.c | 56 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/multihash.c | 166 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/sha1.c | 191 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/sha2big.c | 285 | ||||
-rw-r--r-- | test/monniaux/BearSSL/src/hash/sha2small.c | 341 |
14 files changed, 3072 insertions, 0 deletions
diff --git a/test/monniaux/BearSSL/src/hash/dig_oid.c b/test/monniaux/BearSSL/src/hash/dig_oid.c new file mode 100644 index 00000000..cd9692c9 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/dig_oid.c @@ -0,0 +1,84 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* + * This file contains the encoded OID for the standard hash functions. + * Such OID appear in, for instance, the PKCS#1 v1.5 padding for RSA + * signatures. + */ + +static const unsigned char md5_OID[] = { + 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x02, 0x05 +}; + +static const unsigned char sha1_OID[] = { + 0x2B, 0x0E, 0x03, 0x02, 0x1A +}; + +static const unsigned char sha224_OID[] = { + 0x60, 0x86, 0x48, 0x01, 0x65, 0x03, 0x04, 0x02, 0x04 +}; + +static const unsigned char sha256_OID[] = { + 0x60, 0x86, 0x48, 0x01, 0x65, 0x03, 0x04, 0x02, 0x01 +}; + +static const unsigned char sha384_OID[] = { + 0x60, 0x86, 0x48, 0x01, 0x65, 0x03, 0x04, 0x02, 0x02 +}; + +static const unsigned char sha512_OID[] = { + 0x60, 0x86, 0x48, 0x01, 0x65, 0x03, 0x04, 0x02, 0x03 +}; + +/* see inner.h */ +const unsigned char * +br_digest_OID(int digest_id, size_t *len) +{ + switch (digest_id) { + case br_md5_ID: + *len = sizeof md5_OID; + return md5_OID; + case br_sha1_ID: + *len = sizeof sha1_OID; + return sha1_OID; + case br_sha224_ID: + *len = sizeof sha224_OID; + return sha224_OID; + case br_sha256_ID: + *len = sizeof sha256_OID; + return sha256_OID; + case br_sha384_ID: + *len = sizeof sha384_OID; + return sha384_OID; + case br_sha512_ID: + *len = sizeof sha512_OID; + return sha512_OID; + default: + *len = 0; + return NULL; + } +} diff --git a/test/monniaux/BearSSL/src/hash/dig_size.c b/test/monniaux/BearSSL/src/hash/dig_size.c new file mode 100644 index 00000000..4625d2c6 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/dig_size.c @@ -0,0 +1,50 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* see inner.h */ +size_t +br_digest_size_by_ID(int digest_id) +{ + switch (digest_id) { + case br_md5sha1_ID: + return br_md5_SIZE + br_sha1_SIZE; + case br_md5_ID: + return br_md5_SIZE; + case br_sha1_ID: + return br_sha1_SIZE; + case br_sha224_ID: + return br_sha224_SIZE; + case br_sha256_ID: + return br_sha256_SIZE; + case br_sha384_ID: + return br_sha384_SIZE; + case br_sha512_ID: + return br_sha512_SIZE; + default: + /* abort(); */ + return 0; + } +} diff --git a/test/monniaux/BearSSL/src/hash/ghash_ctmul.c b/test/monniaux/BearSSL/src/hash/ghash_ctmul.c new file mode 100644 index 00000000..36232025 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/ghash_ctmul.c @@ -0,0 +1,345 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* + * We compute "carryless multiplications" through normal integer + * multiplications, masking out enough bits to create "holes" in which + * carries may expand without altering our bits; we really use 8 data + * bits per 32-bit word, spaced every fourth bit. Accumulated carries + * may not exceed 8 in total, which fits in 4 bits. + * + * It would be possible to use a 3-bit spacing, allowing two operands, + * one with 7 non-zero data bits, the other one with 10 or 11 non-zero + * data bits; this asymmetric splitting makes the overall code more + * complex with thresholds and exceptions, and does not appear to be + * worth the effort. + */ + +/* + * We cannot really autodetect whether multiplications are "slow" or + * not. A typical example is the ARM Cortex M0+, which exists in two + * versions: one with a 1-cycle multiplication opcode, the other with + * a 32-cycle multiplication opcode. They both use exactly the same + * architecture and ABI, and cannot be distinguished from each other + * at compile-time. + * + * Since most modern CPU (even embedded CPU) still have fast + * multiplications, we use the "fast mul" code by default. + */ + +#if BR_SLOW_MUL + +/* + * This implementation uses Karatsuba-like reduction to make fewer + * integer multiplications (9 instead of 16), at the expense of extra + * logical operations (XOR, shifts...). On modern x86 CPU that offer + * fast, pipelined multiplications, this code is about twice slower than + * the simpler code with 16 multiplications. This tendency may be + * reversed on low-end platforms with expensive multiplications. + */ + +#define MUL32(h, l, x, y) do { \ + uint64_t mul32tmp = MUL(x, y); \ + (h) = (uint32_t)(mul32tmp >> 32); \ + (l) = (uint32_t)mul32tmp; \ + } while (0) + +static inline void +bmul(uint32_t *hi, uint32_t *lo, uint32_t x, uint32_t y) +{ + uint32_t x0, x1, x2, x3; + uint32_t y0, y1, y2, y3; + uint32_t a0, a1, a2, a3, a4, a5, a6, a7, a8; + uint32_t b0, b1, b2, b3, b4, b5, b6, b7, b8; + + x0 = x & (uint32_t)0x11111111; + x1 = x & (uint32_t)0x22222222; + x2 = x & (uint32_t)0x44444444; + x3 = x & (uint32_t)0x88888888; + y0 = y & (uint32_t)0x11111111; + y1 = y & (uint32_t)0x22222222; + y2 = y & (uint32_t)0x44444444; + y3 = y & (uint32_t)0x88888888; + + /* + * (x0+W*x1)*(y0+W*y1) -> a0:b0 + * (x2+W*x3)*(y2+W*y3) -> a3:b3 + * ((x0+x2)+W*(x1+x3))*((y0+y2)+W*(y1+y3)) -> a6:b6 + */ + a0 = x0; + b0 = y0; + a1 = x1 >> 1; + b1 = y1 >> 1; + a2 = a0 ^ a1; + b2 = b0 ^ b1; + a3 = x2 >> 2; + b3 = y2 >> 2; + a4 = x3 >> 3; + b4 = y3 >> 3; + a5 = a3 ^ a4; + b5 = b3 ^ b4; + a6 = a0 ^ a3; + b6 = b0 ^ b3; + a7 = a1 ^ a4; + b7 = b1 ^ b4; + a8 = a6 ^ a7; + b8 = b6 ^ b7; + + MUL32(b0, a0, b0, a0); + MUL32(b1, a1, b1, a1); + MUL32(b2, a2, b2, a2); + MUL32(b3, a3, b3, a3); + MUL32(b4, a4, b4, a4); + MUL32(b5, a5, b5, a5); + MUL32(b6, a6, b6, a6); + MUL32(b7, a7, b7, a7); + MUL32(b8, a8, b8, a8); + + a0 &= (uint32_t)0x11111111; + a1 &= (uint32_t)0x11111111; + a2 &= (uint32_t)0x11111111; + a3 &= (uint32_t)0x11111111; + a4 &= (uint32_t)0x11111111; + a5 &= (uint32_t)0x11111111; + a6 &= (uint32_t)0x11111111; + a7 &= (uint32_t)0x11111111; + a8 &= (uint32_t)0x11111111; + b0 &= (uint32_t)0x11111111; + b1 &= (uint32_t)0x11111111; + b2 &= (uint32_t)0x11111111; + b3 &= (uint32_t)0x11111111; + b4 &= (uint32_t)0x11111111; + b5 &= (uint32_t)0x11111111; + b6 &= (uint32_t)0x11111111; + b7 &= (uint32_t)0x11111111; + b8 &= (uint32_t)0x11111111; + + a2 ^= a0 ^ a1; + b2 ^= b0 ^ b1; + a0 ^= (a2 << 1) ^ (a1 << 2); + b0 ^= (b2 << 1) ^ (b1 << 2); + a5 ^= a3 ^ a4; + b5 ^= b3 ^ b4; + a3 ^= (a5 << 1) ^ (a4 << 2); + b3 ^= (b5 << 1) ^ (b4 << 2); + a8 ^= a6 ^ a7; + b8 ^= b6 ^ b7; + a6 ^= (a8 << 1) ^ (a7 << 2); + b6 ^= (b8 << 1) ^ (b7 << 2); + a6 ^= a0 ^ a3; + b6 ^= b0 ^ b3; + *lo = a0 ^ (a6 << 2) ^ (a3 << 4); + *hi = b0 ^ (b6 << 2) ^ (b3 << 4) ^ (a6 >> 30) ^ (a3 >> 28); +} + +#else + +/* + * Simple multiplication in GF(2)[X], using 16 integer multiplications. + */ + +static inline void +bmul(uint32_t *hi, uint32_t *lo, uint32_t x, uint32_t y) +{ + uint32_t x0, x1, x2, x3; + uint32_t y0, y1, y2, y3; + uint64_t z0, z1, z2, z3; + uint64_t z; + + x0 = x & (uint32_t)0x11111111; + x1 = x & (uint32_t)0x22222222; + x2 = x & (uint32_t)0x44444444; + x3 = x & (uint32_t)0x88888888; + y0 = y & (uint32_t)0x11111111; + y1 = y & (uint32_t)0x22222222; + y2 = y & (uint32_t)0x44444444; + y3 = y & (uint32_t)0x88888888; + z0 = MUL(x0, y0) ^ MUL(x1, y3) ^ MUL(x2, y2) ^ MUL(x3, y1); + z1 = MUL(x0, y1) ^ MUL(x1, y0) ^ MUL(x2, y3) ^ MUL(x3, y2); + z2 = MUL(x0, y2) ^ MUL(x1, y1) ^ MUL(x2, y0) ^ MUL(x3, y3); + z3 = MUL(x0, y3) ^ MUL(x1, y2) ^ MUL(x2, y1) ^ MUL(x3, y0); + z0 &= (uint64_t)0x1111111111111111; + z1 &= (uint64_t)0x2222222222222222; + z2 &= (uint64_t)0x4444444444444444; + z3 &= (uint64_t)0x8888888888888888; + z = z0 | z1 | z2 | z3; + *lo = (uint32_t)z; + *hi = (uint32_t)(z >> 32); +} + +#endif + +/* see bearssl_hash.h */ +void +br_ghash_ctmul(void *y, const void *h, const void *data, size_t len) +{ + const unsigned char *buf, *hb; + unsigned char *yb; + uint32_t yw[4]; + uint32_t hw[4]; + + /* + * Throughout the loop we handle the y and h values as arrays + * of 32-bit words. + */ + buf = data; + yb = y; + hb = h; + yw[3] = br_dec32be(yb); + yw[2] = br_dec32be(yb + 4); + yw[1] = br_dec32be(yb + 8); + yw[0] = br_dec32be(yb + 12); + hw[3] = br_dec32be(hb); + hw[2] = br_dec32be(hb + 4); + hw[1] = br_dec32be(hb + 8); + hw[0] = br_dec32be(hb + 12); + while (len > 0) { + const unsigned char *src; + unsigned char tmp[16]; + int i; + uint32_t a[9], b[9], zw[8]; + uint32_t c0, c1, c2, c3, d0, d1, d2, d3, e0, e1, e2, e3; + + /* + * Get the next 16-byte block (using zero-padding if + * necessary). + */ + if (len >= 16) { + src = buf; + buf += 16; + len -= 16; + } else { + memcpy(tmp, buf, len); + memset(tmp + len, 0, (sizeof tmp) - len); + src = tmp; + len = 0; + } + + /* + * Decode the block. The GHASH standard mandates + * big-endian encoding. + */ + yw[3] ^= br_dec32be(src); + yw[2] ^= br_dec32be(src + 4); + yw[1] ^= br_dec32be(src + 8); + yw[0] ^= br_dec32be(src + 12); + + /* + * We multiply two 128-bit field elements. We use + * Karatsuba to turn that into three 64-bit + * multiplications, which are themselves done with a + * total of nine 32-bit multiplications. + */ + + /* + * y[0,1]*h[0,1] -> 0..2 + * y[2,3]*h[2,3] -> 3..5 + * (y[0,1]+y[2,3])*(h[0,1]+h[2,3]) -> 6..8 + */ + a[0] = yw[0]; + b[0] = hw[0]; + a[1] = yw[1]; + b[1] = hw[1]; + a[2] = a[0] ^ a[1]; + b[2] = b[0] ^ b[1]; + + a[3] = yw[2]; + b[3] = hw[2]; + a[4] = yw[3]; + b[4] = hw[3]; + a[5] = a[3] ^ a[4]; + b[5] = b[3] ^ b[4]; + + a[6] = a[0] ^ a[3]; + b[6] = b[0] ^ b[3]; + a[7] = a[1] ^ a[4]; + b[7] = b[1] ^ b[4]; + a[8] = a[6] ^ a[7]; + b[8] = b[6] ^ b[7]; + + for (i = 0; i < 9; i ++) { + bmul(&b[i], &a[i], b[i], a[i]); + } + + c0 = a[0]; + c1 = b[0] ^ a[2] ^ a[0] ^ a[1]; + c2 = a[1] ^ b[2] ^ b[0] ^ b[1]; + c3 = b[1]; + d0 = a[3]; + d1 = b[3] ^ a[5] ^ a[3] ^ a[4]; + d2 = a[4] ^ b[5] ^ b[3] ^ b[4]; + d3 = b[4]; + e0 = a[6]; + e1 = b[6] ^ a[8] ^ a[6] ^ a[7]; + e2 = a[7] ^ b[8] ^ b[6] ^ b[7]; + e3 = b[7]; + + e0 ^= c0 ^ d0; + e1 ^= c1 ^ d1; + e2 ^= c2 ^ d2; + e3 ^= c3 ^ d3; + c2 ^= e0; + c3 ^= e1; + d0 ^= e2; + d1 ^= e3; + + /* + * GHASH specification has the bits "reversed" (most + * significant is in fact least significant), which does + * not matter for a carryless multiplication, except that + * the 255-bit result must be shifted by 1 bit. + */ + zw[0] = c0 << 1; + zw[1] = (c1 << 1) | (c0 >> 31); + zw[2] = (c2 << 1) | (c1 >> 31); + zw[3] = (c3 << 1) | (c2 >> 31); + zw[4] = (d0 << 1) | (c3 >> 31); + zw[5] = (d1 << 1) | (d0 >> 31); + zw[6] = (d2 << 1) | (d1 >> 31); + zw[7] = (d3 << 1) | (d2 >> 31); + + /* + * We now do the reduction modulo the field polynomial + * to get back to 128 bits. + */ + for (i = 0; i < 4; i ++) { + uint32_t lw; + + lw = zw[i]; + zw[i + 4] ^= lw ^ (lw >> 1) ^ (lw >> 2) ^ (lw >> 7); + zw[i + 3] ^= (lw << 31) ^ (lw << 30) ^ (lw << 25); + } + memcpy(yw, zw + 4, sizeof yw); + } + + /* + * Encode back the result. + */ + br_enc32be(yb, yw[3]); + br_enc32be(yb + 4, yw[2]); + br_enc32be(yb + 8, yw[1]); + br_enc32be(yb + 12, yw[0]); +} diff --git a/test/monniaux/BearSSL/src/hash/ghash_ctmul32.c b/test/monniaux/BearSSL/src/hash/ghash_ctmul32.c new file mode 100644 index 00000000..c66af465 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/ghash_ctmul32.c @@ -0,0 +1,251 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* + * This implementation uses 32-bit multiplications, and only the low + * 32 bits for each multiplication result. This is meant primarily for + * the ARM Cortex M0 and M0+, whose multiplication opcode does not yield + * the upper 32 bits; but it might also be useful on architectures where + * access to the upper 32 bits requires use of specific registers that + * create contention (e.g. on i386, "mul" necessarily outputs the result + * in edx:eax, while "imul" can use any registers but is limited to the + * low 32 bits). + * + * The implementation trick that is used here is bit-reversing (bit 0 + * is swapped with bit 31, bit 1 with bit 30, and so on). In GF(2)[X], + * for all values x and y, we have: + * rev32(x) * rev32(y) = rev64(x * y) + * In other words, if we bit-reverse (over 32 bits) the operands, then we + * bit-reverse (over 64 bits) the result. + */ + +/* + * Multiplication in GF(2)[X], truncated to its low 32 bits. + */ +static inline uint32_t +bmul32(uint32_t x, uint32_t y) +{ + uint32_t x0, x1, x2, x3; + uint32_t y0, y1, y2, y3; + uint32_t z0, z1, z2, z3; + + x0 = x & (uint32_t)0x11111111; + x1 = x & (uint32_t)0x22222222; + x2 = x & (uint32_t)0x44444444; + x3 = x & (uint32_t)0x88888888; + y0 = y & (uint32_t)0x11111111; + y1 = y & (uint32_t)0x22222222; + y2 = y & (uint32_t)0x44444444; + y3 = y & (uint32_t)0x88888888; + z0 = (x0 * y0) ^ (x1 * y3) ^ (x2 * y2) ^ (x3 * y1); + z1 = (x0 * y1) ^ (x1 * y0) ^ (x2 * y3) ^ (x3 * y2); + z2 = (x0 * y2) ^ (x1 * y1) ^ (x2 * y0) ^ (x3 * y3); + z3 = (x0 * y3) ^ (x1 * y2) ^ (x2 * y1) ^ (x3 * y0); + z0 &= (uint32_t)0x11111111; + z1 &= (uint32_t)0x22222222; + z2 &= (uint32_t)0x44444444; + z3 &= (uint32_t)0x88888888; + return z0 | z1 | z2 | z3; +} + +/* + * Bit-reverse a 32-bit word. + */ +static uint32_t +rev32(uint32_t x) +{ +#define RMS(m, s) do { \ + x = ((x & (uint32_t)(m)) << (s)) \ + | ((x >> (s)) & (uint32_t)(m)); \ + } while (0) + + RMS(0x55555555, 1); + RMS(0x33333333, 2); + RMS(0x0F0F0F0F, 4); + RMS(0x00FF00FF, 8); + return (x << 16) | (x >> 16); + +#undef RMS +} + +/* see bearssl_hash.h */ +void +br_ghash_ctmul32(void *y, const void *h, const void *data, size_t len) +{ + /* + * This implementation is similar to br_ghash_ctmul() except + * that we have to do the multiplication twice, with the + * "normal" and "bit reversed" operands. Hence we end up with + * eighteen 32-bit multiplications instead of nine. + */ + + const unsigned char *buf, *hb; + unsigned char *yb; + uint32_t yw[4]; + uint32_t hw[4], hwr[4]; + + buf = data; + yb = y; + hb = h; + yw[3] = br_dec32be(yb); + yw[2] = br_dec32be(yb + 4); + yw[1] = br_dec32be(yb + 8); + yw[0] = br_dec32be(yb + 12); + hw[3] = br_dec32be(hb); + hw[2] = br_dec32be(hb + 4); + hw[1] = br_dec32be(hb + 8); + hw[0] = br_dec32be(hb + 12); + hwr[3] = rev32(hw[3]); + hwr[2] = rev32(hw[2]); + hwr[1] = rev32(hw[1]); + hwr[0] = rev32(hw[0]); + while (len > 0) { + const unsigned char *src; + unsigned char tmp[16]; + int i; + uint32_t a[18], b[18], c[18]; + uint32_t d0, d1, d2, d3, d4, d5, d6, d7; + uint32_t zw[8]; + + if (len >= 16) { + src = buf; + buf += 16; + len -= 16; + } else { + memcpy(tmp, buf, len); + memset(tmp + len, 0, (sizeof tmp) - len); + src = tmp; + len = 0; + } + yw[3] ^= br_dec32be(src); + yw[2] ^= br_dec32be(src + 4); + yw[1] ^= br_dec32be(src + 8); + yw[0] ^= br_dec32be(src + 12); + + /* + * We are using Karatsuba: the 128x128 multiplication is + * reduced to three 64x64 multiplications, hence nine + * 32x32 multiplications. With the bit-reversal trick, + * we have to perform 18 32x32 multiplications. + */ + + /* + * y[0,1]*h[0,1] -> 0,1,4 + * y[2,3]*h[2,3] -> 2,3,5 + * (y[0,1]+y[2,3])*(h[0,1]+h[2,3]) -> 6,7,8 + */ + + a[0] = yw[0]; + a[1] = yw[1]; + a[2] = yw[2]; + a[3] = yw[3]; + a[4] = a[0] ^ a[1]; + a[5] = a[2] ^ a[3]; + a[6] = a[0] ^ a[2]; + a[7] = a[1] ^ a[3]; + a[8] = a[6] ^ a[7]; + + a[ 9] = rev32(yw[0]); + a[10] = rev32(yw[1]); + a[11] = rev32(yw[2]); + a[12] = rev32(yw[3]); + a[13] = a[ 9] ^ a[10]; + a[14] = a[11] ^ a[12]; + a[15] = a[ 9] ^ a[11]; + a[16] = a[10] ^ a[12]; + a[17] = a[15] ^ a[16]; + + b[0] = hw[0]; + b[1] = hw[1]; + b[2] = hw[2]; + b[3] = hw[3]; + b[4] = b[0] ^ b[1]; + b[5] = b[2] ^ b[3]; + b[6] = b[0] ^ b[2]; + b[7] = b[1] ^ b[3]; + b[8] = b[6] ^ b[7]; + + b[ 9] = hwr[0]; + b[10] = hwr[1]; + b[11] = hwr[2]; + b[12] = hwr[3]; + b[13] = b[ 9] ^ b[10]; + b[14] = b[11] ^ b[12]; + b[15] = b[ 9] ^ b[11]; + b[16] = b[10] ^ b[12]; + b[17] = b[15] ^ b[16]; + + for (i = 0; i < 18; i ++) { + c[i] = bmul32(a[i], b[i]); + } + + c[4] ^= c[0] ^ c[1]; + c[5] ^= c[2] ^ c[3]; + c[8] ^= c[6] ^ c[7]; + + c[13] ^= c[ 9] ^ c[10]; + c[14] ^= c[11] ^ c[12]; + c[17] ^= c[15] ^ c[16]; + + /* + * y[0,1]*h[0,1] -> 0,9^4,1^13,10 + * y[2,3]*h[2,3] -> 2,11^5,3^14,12 + * (y[0,1]+y[2,3])*(h[0,1]+h[2,3]) -> 6,15^8,7^17,16 + */ + d0 = c[0]; + d1 = c[4] ^ (rev32(c[9]) >> 1); + d2 = c[1] ^ c[0] ^ c[2] ^ c[6] ^ (rev32(c[13]) >> 1); + d3 = c[4] ^ c[5] ^ c[8] + ^ (rev32(c[10] ^ c[9] ^ c[11] ^ c[15]) >> 1); + d4 = c[2] ^ c[1] ^ c[3] ^ c[7] + ^ (rev32(c[13] ^ c[14] ^ c[17]) >> 1); + d5 = c[5] ^ (rev32(c[11] ^ c[10] ^ c[12] ^ c[16]) >> 1); + d6 = c[3] ^ (rev32(c[14]) >> 1); + d7 = rev32(c[12]) >> 1; + + zw[0] = d0 << 1; + zw[1] = (d1 << 1) | (d0 >> 31); + zw[2] = (d2 << 1) | (d1 >> 31); + zw[3] = (d3 << 1) | (d2 >> 31); + zw[4] = (d4 << 1) | (d3 >> 31); + zw[5] = (d5 << 1) | (d4 >> 31); + zw[6] = (d6 << 1) | (d5 >> 31); + zw[7] = (d7 << 1) | (d6 >> 31); + + for (i = 0; i < 4; i ++) { + uint32_t lw; + + lw = zw[i]; + zw[i + 4] ^= lw ^ (lw >> 1) ^ (lw >> 2) ^ (lw >> 7); + zw[i + 3] ^= (lw << 31) ^ (lw << 30) ^ (lw << 25); + } + memcpy(yw, zw + 4, sizeof yw); + } + br_enc32be(yb, yw[3]); + br_enc32be(yb + 4, yw[2]); + br_enc32be(yb + 8, yw[1]); + br_enc32be(yb + 12, yw[0]); +} diff --git a/test/monniaux/BearSSL/src/hash/ghash_ctmul64.c b/test/monniaux/BearSSL/src/hash/ghash_ctmul64.c new file mode 100644 index 00000000..a46f16fe --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/ghash_ctmul64.c @@ -0,0 +1,154 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* + * This is the 64-bit variant of br_ghash_ctmul32(), with 64-bit operands + * and bit reversal of 64-bit words. + */ + +static inline uint64_t +bmul64(uint64_t x, uint64_t y) +{ + uint64_t x0, x1, x2, x3; + uint64_t y0, y1, y2, y3; + uint64_t z0, z1, z2, z3; + + x0 = x & (uint64_t)0x1111111111111111; + x1 = x & (uint64_t)0x2222222222222222; + x2 = x & (uint64_t)0x4444444444444444; + x3 = x & (uint64_t)0x8888888888888888; + y0 = y & (uint64_t)0x1111111111111111; + y1 = y & (uint64_t)0x2222222222222222; + y2 = y & (uint64_t)0x4444444444444444; + y3 = y & (uint64_t)0x8888888888888888; + z0 = (x0 * y0) ^ (x1 * y3) ^ (x2 * y2) ^ (x3 * y1); + z1 = (x0 * y1) ^ (x1 * y0) ^ (x2 * y3) ^ (x3 * y2); + z2 = (x0 * y2) ^ (x1 * y1) ^ (x2 * y0) ^ (x3 * y3); + z3 = (x0 * y3) ^ (x1 * y2) ^ (x2 * y1) ^ (x3 * y0); + z0 &= (uint64_t)0x1111111111111111; + z1 &= (uint64_t)0x2222222222222222; + z2 &= (uint64_t)0x4444444444444444; + z3 &= (uint64_t)0x8888888888888888; + return z0 | z1 | z2 | z3; +} + +static uint64_t +rev64(uint64_t x) +{ +#define RMS(m, s) do { \ + x = ((x & (uint64_t)(m)) << (s)) \ + | ((x >> (s)) & (uint64_t)(m)); \ + } while (0) + + RMS(0x5555555555555555, 1); + RMS(0x3333333333333333, 2); + RMS(0x0F0F0F0F0F0F0F0F, 4); + RMS(0x00FF00FF00FF00FF, 8); + RMS(0x0000FFFF0000FFFF, 16); + return (x << 32) | (x >> 32); + +#undef RMS +} + +/* see bearssl_ghash.h */ +void +br_ghash_ctmul64(void *y, const void *h, const void *data, size_t len) +{ + const unsigned char *buf, *hb; + unsigned char *yb; + uint64_t y0, y1; + uint64_t h0, h1, h2, h0r, h1r, h2r; + + buf = data; + yb = y; + hb = h; + y1 = br_dec64be(yb); + y0 = br_dec64be(yb + 8); + h1 = br_dec64be(hb); + h0 = br_dec64be(hb + 8); + h0r = rev64(h0); + h1r = rev64(h1); + h2 = h0 ^ h1; + h2r = h0r ^ h1r; + while (len > 0) { + const unsigned char *src; + unsigned char tmp[16]; + uint64_t y0r, y1r, y2, y2r; + uint64_t z0, z1, z2, z0h, z1h, z2h; + uint64_t v0, v1, v2, v3; + + if (len >= 16) { + src = buf; + buf += 16; + len -= 16; + } else { + memcpy(tmp, buf, len); + memset(tmp + len, 0, (sizeof tmp) - len); + src = tmp; + len = 0; + } + y1 ^= br_dec64be(src); + y0 ^= br_dec64be(src + 8); + + y0r = rev64(y0); + y1r = rev64(y1); + y2 = y0 ^ y1; + y2r = y0r ^ y1r; + + z0 = bmul64(y0, h0); + z1 = bmul64(y1, h1); + z2 = bmul64(y2, h2); + z0h = bmul64(y0r, h0r); + z1h = bmul64(y1r, h1r); + z2h = bmul64(y2r, h2r); + z2 ^= z0 ^ z1; + z2h ^= z0h ^ z1h; + z0h = rev64(z0h) >> 1; + z1h = rev64(z1h) >> 1; + z2h = rev64(z2h) >> 1; + + v0 = z0; + v1 = z0h ^ z2; + v2 = z1 ^ z2h; + v3 = z1h; + + v3 = (v3 << 1) | (v2 >> 63); + v2 = (v2 << 1) | (v1 >> 63); + v1 = (v1 << 1) | (v0 >> 63); + v0 = (v0 << 1); + + v2 ^= v0 ^ (v0 >> 1) ^ (v0 >> 2) ^ (v0 >> 7); + v1 ^= (v0 << 63) ^ (v0 << 62) ^ (v0 << 57); + v3 ^= v1 ^ (v1 >> 1) ^ (v1 >> 2) ^ (v1 >> 7); + v2 ^= (v1 << 63) ^ (v1 << 62) ^ (v1 << 57); + + y0 = v2; + y1 = v3; + } + + br_enc64be(yb, y1); + br_enc64be(yb + 8, y0); +} diff --git a/test/monniaux/BearSSL/src/hash/ghash_pclmul.c b/test/monniaux/BearSSL/src/hash/ghash_pclmul.c new file mode 100644 index 00000000..a58e7dc0 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/ghash_pclmul.c @@ -0,0 +1,389 @@ +/* + * Copyright (c) 2017 Thomas Pornin <pornin@bolet.org> + * + * 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. + */ + +#define BR_ENABLE_INTRINSICS 1 +#include "inner.h" + +/* + * This is the GHASH implementation that leverages the pclmulqdq opcode + * (from the AES-NI instructions). + */ + +#if BR_AES_X86NI + +/* + * Test CPU support for PCLMULQDQ. + */ +static inline int +pclmul_supported(void) +{ + /* + * Bit mask for features in ECX: + * 1 PCLMULQDQ support + */ + return br_cpuid(0, 0, 0x00000002, 0); +} + +/* see bearssl_hash.h */ +br_ghash +br_ghash_pclmul_get(void) +{ + return pclmul_supported() ? &br_ghash_pclmul : 0; +} + +BR_TARGETS_X86_UP + +/* + * GHASH is defined over elements of GF(2^128) with "full little-endian" + * representation: leftmost byte is least significant, and, within each + * byte, leftmost _bit_ is least significant. The natural ordering in + * x86 is "mixed little-endian": bytes are ordered from least to most + * significant, but bits within a byte are in most-to-least significant + * order. Going to full little-endian representation would require + * reversing bits within each byte, which is doable but expensive. + * + * Instead, we go to full big-endian representation, by swapping bytes + * around, which is done with a single _mm_shuffle_epi8() opcode (it + * comes with SSSE3; all CPU that offer pclmulqdq also have SSSE3). We + * can use a full big-endian representation because in a carryless + * multiplication, we have a nice bit reversal property: + * + * rev_128(x) * rev_128(y) = rev_255(x * y) + * + * So by using full big-endian, we still get the right result, except + * that it is right-shifted by 1 bit. The left-shift is relatively + * inexpensive, and it can be mutualised. + * + * + * Since SSE2 opcodes do not have facilities for shitfting full 128-bit + * values with bit precision, we have to break down values into 64-bit + * chunks. We number chunks from 0 to 3 in left to right order. + */ + +/* + * Byte-swap a complete 128-bit value. This normally uses + * _mm_shuffle_epi8(), which gets translated to pshufb (an SSSE3 opcode). + * However, this crashes old Clang versions, so, for Clang before 3.8, + * we use an alternate (and less efficient) version. + */ +#if BR_CLANG && !BR_CLANG_3_8 +#define BYTESWAP_DECL +#define BYTESWAP_PREP (void)0 +#define BYTESWAP(x) do { \ + __m128i byteswap1, byteswap2; \ + byteswap1 = (x); \ + byteswap2 = _mm_srli_epi16(byteswap1, 8); \ + byteswap1 = _mm_slli_epi16(byteswap1, 8); \ + byteswap1 = _mm_or_si128(byteswap1, byteswap2); \ + byteswap1 = _mm_shufflelo_epi16(byteswap1, 0x1B); \ + byteswap1 = _mm_shufflehi_epi16(byteswap1, 0x1B); \ + (x) = _mm_shuffle_epi32(byteswap1, 0x4E); \ + } while (0) +#else +#define BYTESWAP_DECL __m128i byteswap_index; +#define BYTESWAP_PREP do { \ + byteswap_index = _mm_set_epi8( \ + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15); \ + } while (0) +#define BYTESWAP(x) do { \ + (x) = _mm_shuffle_epi8((x), byteswap_index); \ + } while (0) +#endif + +/* + * Call pclmulqdq. Clang appears to have trouble with the intrinsic, so, + * for that compiler, we use inline assembly. Inline assembly is + * potentially a bit slower because the compiler does not understand + * what the opcode does, and thus cannot optimize instruction + * scheduling. + * + * We use a target of "sse2" only, so that Clang may still handle the + * '__m128i' type and allocate SSE2 registers. + */ +#if BR_CLANG +BR_TARGET("sse2") +static inline __m128i +pclmulqdq00(__m128i x, __m128i y) +{ + __asm__ ("pclmulqdq $0x00, %1, %0" : "+x" (x) : "x" (y)); + return x; +} +BR_TARGET("sse2") +static inline __m128i +pclmulqdq11(__m128i x, __m128i y) +{ + __asm__ ("pclmulqdq $0x11, %1, %0" : "+x" (x) : "x" (y)); + return x; +} +#else +#define pclmulqdq00(x, y) _mm_clmulepi64_si128(x, y, 0x00) +#define pclmulqdq11(x, y) _mm_clmulepi64_si128(x, y, 0x11) +#endif + +/* + * From a 128-bit value kw, compute kx as the XOR of the two 64-bit + * halves of kw (into the right half of kx; left half is unspecified). + */ +#define BK(kw, kx) do { \ + kx = _mm_xor_si128(kw, _mm_shuffle_epi32(kw, 0x0E)); \ + } while (0) + +/* + * Combine two 64-bit values (k0:k1) into a 128-bit (kw) value and + * the XOR of the two values (kx). + */ +#define PBK(k0, k1, kw, kx) do { \ + kw = _mm_unpacklo_epi64(k1, k0); \ + kx = _mm_xor_si128(k0, k1); \ + } while (0) + +/* + * Left-shift by 1 bit a 256-bit value (in four 64-bit words). + */ +#define SL_256(x0, x1, x2, x3) do { \ + x0 = _mm_or_si128( \ + _mm_slli_epi64(x0, 1), \ + _mm_srli_epi64(x1, 63)); \ + x1 = _mm_or_si128( \ + _mm_slli_epi64(x1, 1), \ + _mm_srli_epi64(x2, 63)); \ + x2 = _mm_or_si128( \ + _mm_slli_epi64(x2, 1), \ + _mm_srli_epi64(x3, 63)); \ + x3 = _mm_slli_epi64(x3, 1); \ + } while (0) + +/* + * Perform reduction in GF(2^128). The 256-bit value is in x0..x3; + * result is written in x0..x1. + */ +#define REDUCE_F128(x0, x1, x2, x3) do { \ + x1 = _mm_xor_si128( \ + x1, \ + _mm_xor_si128( \ + _mm_xor_si128( \ + x3, \ + _mm_srli_epi64(x3, 1)), \ + _mm_xor_si128( \ + _mm_srli_epi64(x3, 2), \ + _mm_srli_epi64(x3, 7)))); \ + x2 = _mm_xor_si128( \ + _mm_xor_si128( \ + x2, \ + _mm_slli_epi64(x3, 63)), \ + _mm_xor_si128( \ + _mm_slli_epi64(x3, 62), \ + _mm_slli_epi64(x3, 57))); \ + x0 = _mm_xor_si128( \ + x0, \ + _mm_xor_si128( \ + _mm_xor_si128( \ + x2, \ + _mm_srli_epi64(x2, 1)), \ + _mm_xor_si128( \ + _mm_srli_epi64(x2, 2), \ + _mm_srli_epi64(x2, 7)))); \ + x1 = _mm_xor_si128( \ + _mm_xor_si128( \ + x1, \ + _mm_slli_epi64(x2, 63)), \ + _mm_xor_si128( \ + _mm_slli_epi64(x2, 62), \ + _mm_slli_epi64(x2, 57))); \ + } while (0) + +/* + * Square value kw into (dw,dx). + */ +#define SQUARE_F128(kw, dw, dx) do { \ + __m128i z0, z1, z2, z3; \ + z1 = pclmulqdq11(kw, kw); \ + z3 = pclmulqdq00(kw, kw); \ + z0 = _mm_shuffle_epi32(z1, 0x0E); \ + z2 = _mm_shuffle_epi32(z3, 0x0E); \ + SL_256(z0, z1, z2, z3); \ + REDUCE_F128(z0, z1, z2, z3); \ + PBK(z0, z1, dw, dx); \ + } while (0) + +/* see bearssl_hash.h */ +BR_TARGET("ssse3,pclmul") +void +br_ghash_pclmul(void *y, const void *h, const void *data, size_t len) +{ + const unsigned char *buf1, *buf2; + unsigned char tmp[64]; + size_t num4, num1; + __m128i yw, h1w, h1x; + BYTESWAP_DECL + + /* + * We split data into two chunks. First chunk starts at buf1 + * and contains num4 blocks of 64-byte values. Second chunk + * starts at buf2 and contains num1 blocks of 16-byte values. + * We want the first chunk to be as large as possible. + */ + buf1 = data; + num4 = len >> 6; + len &= 63; + buf2 = buf1 + (num4 << 6); + num1 = (len + 15) >> 4; + if ((len & 15) != 0) { + memcpy(tmp, buf2, len); + memset(tmp + len, 0, (num1 << 4) - len); + buf2 = tmp; + } + + /* + * Preparatory step for endian conversions. + */ + BYTESWAP_PREP; + + /* + * Load y and h. + */ + yw = _mm_loadu_si128(y); + h1w = _mm_loadu_si128(h); + BYTESWAP(yw); + BYTESWAP(h1w); + BK(h1w, h1x); + + if (num4 > 0) { + __m128i h2w, h2x, h3w, h3x, h4w, h4x; + __m128i t0, t1, t2, t3; + + /* + * Compute h2 = h^2. + */ + SQUARE_F128(h1w, h2w, h2x); + + /* + * Compute h3 = h^3 = h*(h^2). + */ + t1 = pclmulqdq11(h1w, h2w); + t3 = pclmulqdq00(h1w, h2w); + t2 = _mm_xor_si128(pclmulqdq00(h1x, h2x), + _mm_xor_si128(t1, t3)); + t0 = _mm_shuffle_epi32(t1, 0x0E); + t1 = _mm_xor_si128(t1, _mm_shuffle_epi32(t2, 0x0E)); + t2 = _mm_xor_si128(t2, _mm_shuffle_epi32(t3, 0x0E)); + SL_256(t0, t1, t2, t3); + REDUCE_F128(t0, t1, t2, t3); + PBK(t0, t1, h3w, h3x); + + /* + * Compute h4 = h^4 = (h^2)^2. + */ + SQUARE_F128(h2w, h4w, h4x); + + while (num4 -- > 0) { + __m128i aw0, aw1, aw2, aw3; + __m128i ax0, ax1, ax2, ax3; + + aw0 = _mm_loadu_si128((void *)(buf1 + 0)); + aw1 = _mm_loadu_si128((void *)(buf1 + 16)); + aw2 = _mm_loadu_si128((void *)(buf1 + 32)); + aw3 = _mm_loadu_si128((void *)(buf1 + 48)); + BYTESWAP(aw0); + BYTESWAP(aw1); + BYTESWAP(aw2); + BYTESWAP(aw3); + buf1 += 64; + + aw0 = _mm_xor_si128(aw0, yw); + BK(aw1, ax1); + BK(aw2, ax2); + BK(aw3, ax3); + BK(aw0, ax0); + + t1 = _mm_xor_si128( + _mm_xor_si128( + pclmulqdq11(aw0, h4w), + pclmulqdq11(aw1, h3w)), + _mm_xor_si128( + pclmulqdq11(aw2, h2w), + pclmulqdq11(aw3, h1w))); + t3 = _mm_xor_si128( + _mm_xor_si128( + pclmulqdq00(aw0, h4w), + pclmulqdq00(aw1, h3w)), + _mm_xor_si128( + pclmulqdq00(aw2, h2w), + pclmulqdq00(aw3, h1w))); + t2 = _mm_xor_si128( + _mm_xor_si128( + pclmulqdq00(ax0, h4x), + pclmulqdq00(ax1, h3x)), + _mm_xor_si128( + pclmulqdq00(ax2, h2x), + pclmulqdq00(ax3, h1x))); + t2 = _mm_xor_si128(t2, _mm_xor_si128(t1, t3)); + t0 = _mm_shuffle_epi32(t1, 0x0E); + t1 = _mm_xor_si128(t1, _mm_shuffle_epi32(t2, 0x0E)); + t2 = _mm_xor_si128(t2, _mm_shuffle_epi32(t3, 0x0E)); + SL_256(t0, t1, t2, t3); + REDUCE_F128(t0, t1, t2, t3); + yw = _mm_unpacklo_epi64(t1, t0); + } + } + + while (num1 -- > 0) { + __m128i aw, ax; + __m128i t0, t1, t2, t3; + + aw = _mm_loadu_si128((void *)buf2); + BYTESWAP(aw); + buf2 += 16; + + aw = _mm_xor_si128(aw, yw); + BK(aw, ax); + + t1 = pclmulqdq11(aw, h1w); + t3 = pclmulqdq00(aw, h1w); + t2 = pclmulqdq00(ax, h1x); + t2 = _mm_xor_si128(t2, _mm_xor_si128(t1, t3)); + t0 = _mm_shuffle_epi32(t1, 0x0E); + t1 = _mm_xor_si128(t1, _mm_shuffle_epi32(t2, 0x0E)); + t2 = _mm_xor_si128(t2, _mm_shuffle_epi32(t3, 0x0E)); + SL_256(t0, t1, t2, t3); + REDUCE_F128(t0, t1, t2, t3); + yw = _mm_unpacklo_epi64(t1, t0); + } + + BYTESWAP(yw); + _mm_storeu_si128(y, yw); +} + +BR_TARGETS_X86_DOWN + +#else + +/* see bearssl_hash.h */ +br_ghash +br_ghash_pclmul_get(void) +{ + return 0; +} + +#endif diff --git a/test/monniaux/BearSSL/src/hash/ghash_pwr8.c b/test/monniaux/BearSSL/src/hash/ghash_pwr8.c new file mode 100644 index 00000000..2e7b0f4c --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/ghash_pwr8.c @@ -0,0 +1,411 @@ +/* + * Copyright (c) 2017 Thomas Pornin <pornin@bolet.org> + * + * 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. + */ + +#define BR_POWER_ASM_MACROS 1 +#include "inner.h" + +/* + * This is the GHASH implementation that leverages the POWER8 opcodes. + */ + +#if BR_POWER8 + +/* + * Some symbolic names for registers. + * HB0 = 16 bytes of value 0 + * HB1 = 16 bytes of value 1 + * HB2 = 16 bytes of value 2 + * HB6 = 16 bytes of value 6 + * HB7 = 16 bytes of value 7 + * TT0, TT1 and TT2 are temporaries + * + * BSW holds the pattern for byteswapping 32-bit words; this is set only + * on little-endian systems. XBSW is the same register with the +32 offset + * for access with the VSX opcodes. + */ +#define HB0 0 +#define HB1 1 +#define HB2 2 +#define HB6 3 +#define HB7 4 +#define TT0 5 +#define TT1 6 +#define TT2 7 + +#define BSW 8 +#define XBSW 40 + +/* + * Macro to initialise the constants. + */ +#define INIT \ + vxor(HB0, HB0, HB0) \ + vspltisb(HB1, 1) \ + vspltisb(HB2, 2) \ + vspltisb(HB6, 6) \ + vspltisb(HB7, 7) \ + INIT_BSW + +/* + * Fix endianness of a value after reading it or before writing it, if + * necessary. + */ +#if BR_POWER8_LE +#define INIT_BSW lxvw4x(XBSW, 0, %[idx2be]) +#define FIX_ENDIAN(xx) vperm(xx, xx, xx, BSW) +#else +#define INIT_BSW +#define FIX_ENDIAN(xx) +#endif + +/* + * Left-shift x0:x1 by one bit to the left. This is a corrective action + * needed because GHASH is defined in full little-endian specification, + * while the opcodes use full big-endian convention, so the 255-bit product + * ends up one bit to the right. + */ +#define SL_256(x0, x1) \ + vsldoi(TT0, HB0, x1, 1) \ + vsl(x0, x0, HB1) \ + vsr(TT0, TT0, HB7) \ + vsl(x1, x1, HB1) \ + vxor(x0, x0, TT0) + +/* + * Reduce x0:x1 in GF(2^128), result in xd (register xd may be the same as + * x0 or x1, or a different register). x0 and x1 are modified. + */ +#define REDUCE_F128(xd, x0, x1) \ + vxor(x0, x0, x1) \ + vsr(TT0, x1, HB1) \ + vsr(TT1, x1, HB2) \ + vsr(TT2, x1, HB7) \ + vxor(x0, x0, TT0) \ + vxor(TT1, TT1, TT2) \ + vxor(x0, x0, TT1) \ + vsldoi(x1, x1, HB0, 15) \ + vsl(TT1, x1, HB6) \ + vsl(TT2, x1, HB1) \ + vxor(x1, TT1, TT2) \ + vsr(TT0, x1, HB1) \ + vsr(TT1, x1, HB2) \ + vsr(TT2, x1, HB7) \ + vxor(x0, x0, x1) \ + vxor(x0, x0, TT0) \ + vxor(TT1, TT1, TT2) \ + vxor(xd, x0, TT1) + +/* see bearssl_hash.h */ +void +br_ghash_pwr8(void *y, const void *h, const void *data, size_t len) +{ + const unsigned char *buf1, *buf2; + size_t num4, num1; + unsigned char tmp[64]; + long cc0, cc1, cc2, cc3; + +#if BR_POWER8_LE + static const uint32_t idx2be[] = { + 0x03020100, 0x07060504, 0x0B0A0908, 0x0F0E0D0C + }; +#endif + + buf1 = data; + + /* + * Assembly code requires data into two chunks; first chunk + * must contain a number of blocks which is a multiple of 4. + * Since the processing for the first chunk is faster, we want + * to make it as big as possible. + * + * For the remainder, there are two possibilities: + * -- if the remainder size is a multiple of 16, then use it + * in place; + * -- otherwise, copy it to the tmp[] array and pad it with + * zeros. + */ + num4 = len >> 6; + buf2 = buf1 + (num4 << 6); + len &= 63; + num1 = (len + 15) >> 4; + if ((len & 15) != 0) { + memcpy(tmp, buf2, len); + memset(tmp + len, 0, (num1 << 4) - len); + buf2 = tmp; + } + + cc0 = 0; + cc1 = 16; + cc2 = 32; + cc3 = 48; + asm volatile ( + INIT + + /* + * Load current h (denoted hereafter h1) in v9. + */ + lxvw4x(41, 0, %[h]) + FIX_ENDIAN(9) + + /* + * Load current y into v28. + */ + lxvw4x(60, 0, %[y]) + FIX_ENDIAN(28) + + /* + * Split h1 into three registers: + * v17 = h1_1:h1_0 + * v18 = 0:h1_0 + * v19 = h1_1:0 + */ + xxpermdi(49, 41, 41, 2) + vsldoi(18, HB0, 9, 8) + vsldoi(19, 9, HB0, 8) + + /* + * If num4 is 0, skip directly to the second chunk. + */ + cmpldi(%[num4], 0) + beq(chunk1) + + /* + * Compute h2 = h*h in v10. + */ + vpmsumd(10, 18, 18) + vpmsumd(11, 19, 19) + SL_256(10, 11) + REDUCE_F128(10, 10, 11) + + /* + * Compute h3 = h*h*h in v11. + * We first split h2 into: + * v10 = h2_0:h2_1 + * v11 = 0:h2_0 + * v12 = h2_1:0 + * Then we do the product with h1, and reduce into v11. + */ + vsldoi(11, HB0, 10, 8) + vsldoi(12, 10, HB0, 8) + vpmsumd(13, 10, 17) + vpmsumd(11, 11, 18) + vpmsumd(12, 12, 19) + vsldoi(14, HB0, 13, 8) + vsldoi(15, 13, HB0, 8) + vxor(11, 11, 14) + vxor(12, 12, 15) + SL_256(11, 12) + REDUCE_F128(11, 11, 12) + + /* + * Compute h4 = h*h*h*h in v12. This is done by squaring h2. + */ + vsldoi(12, HB0, 10, 8) + vsldoi(13, 10, HB0, 8) + vpmsumd(12, 12, 12) + vpmsumd(13, 13, 13) + SL_256(12, 13) + REDUCE_F128(12, 12, 13) + + /* + * Repack h1, h2, h3 and h4: + * v13 = h4_0:h3_0 + * v14 = h4_1:h3_1 + * v15 = h2_0:h1_0 + * v16 = h2_1:h1_1 + */ + xxpermdi(45, 44, 43, 0) + xxpermdi(46, 44, 43, 3) + xxpermdi(47, 42, 41, 0) + xxpermdi(48, 42, 41, 3) + + /* + * Loop for each group of four blocks. + */ + mtctr(%[num4]) + label(loop4) + /* + * Read the four next blocks. + * v20 = y + a0 = b0 + * v21 = a1 = b1 + * v22 = a2 = b2 + * v23 = a3 = b3 + */ + lxvw4x(52, %[cc0], %[buf1]) + lxvw4x(53, %[cc1], %[buf1]) + lxvw4x(54, %[cc2], %[buf1]) + lxvw4x(55, %[cc3], %[buf1]) + FIX_ENDIAN(20) + FIX_ENDIAN(21) + FIX_ENDIAN(22) + FIX_ENDIAN(23) + addi(%[buf1], %[buf1], 64) + vxor(20, 20, 28) + + /* + * Repack the blocks into v9, v10, v11 and v12. + * v9 = b0_0:b1_0 + * v10 = b0_1:b1_1 + * v11 = b2_0:b3_0 + * v12 = b2_1:b3_1 + */ + xxpermdi(41, 52, 53, 0) + xxpermdi(42, 52, 53, 3) + xxpermdi(43, 54, 55, 0) + xxpermdi(44, 54, 55, 3) + + /* + * Compute the products. + * v20 = b0_0*h4_0 + b1_0*h3_0 + * v21 = b0_1*h4_0 + b1_1*h3_0 + * v22 = b0_0*h4_1 + b1_0*h3_1 + * v23 = b0_1*h4_1 + b1_1*h3_1 + * v24 = b2_0*h2_0 + b3_0*h1_0 + * v25 = b2_1*h2_0 + b3_1*h1_0 + * v26 = b2_0*h2_1 + b3_0*h1_1 + * v27 = b2_1*h2_1 + b3_1*h1_1 + */ + vpmsumd(20, 13, 9) + vpmsumd(21, 13, 10) + vpmsumd(22, 14, 9) + vpmsumd(23, 14, 10) + vpmsumd(24, 15, 11) + vpmsumd(25, 15, 12) + vpmsumd(26, 16, 11) + vpmsumd(27, 16, 12) + + /* + * Sum products into a single 256-bit result in v11:v12. + */ + vxor(11, 20, 24) + vxor(12, 23, 27) + vxor( 9, 21, 22) + vxor(10, 25, 26) + vxor(20, 9, 10) + vsldoi( 9, HB0, 20, 8) + vsldoi(10, 20, HB0, 8) + vxor(11, 11, 9) + vxor(12, 12, 10) + + /* + * Fix and reduce in GF(2^128); this is the new y (in v28). + */ + SL_256(11, 12) + REDUCE_F128(28, 11, 12) + + /* + * Loop for next group of four blocks. + */ + bdnz(loop4) + + /* + * Process second chunk, one block at a time. + */ + label(chunk1) + cmpldi(%[num1], 0) + beq(done) + + mtctr(%[num1]) + label(loop1) + /* + * Load next data block and XOR it into y. + */ + lxvw4x(41, 0, %[buf2]) +#if BR_POWER8_LE + FIX_ENDIAN(9) +#endif + addi(%[buf2], %[buf2], 16) + vxor(9, 28, 9) + + /* + * Split y into doublewords: + * v9 = y_0:y_1 + * v10 = 0:y_0 + * v11 = y_1:0 + */ + vsldoi(10, HB0, 9, 8) + vsldoi(11, 9, HB0, 8) + + /* + * Compute products with h: + * v12 = y_0 * h_0 + * v13 = y_1 * h_1 + * v14 = y_1 * h_0 + y_0 * h_1 + */ + vpmsumd(14, 9, 17) + vpmsumd(12, 10, 18) + vpmsumd(13, 11, 19) + + /* + * Propagate v14 into v12:v13 to finalise product. + */ + vsldoi(10, HB0, 14, 8) + vsldoi(11, 14, HB0, 8) + vxor(12, 12, 10) + vxor(13, 13, 11) + + /* + * Fix result and reduce into v28 (next value for y). + */ + SL_256(12, 13) + REDUCE_F128(28, 12, 13) + bdnz(loop1) + + label(done) + /* + * Write back the new y. + */ + FIX_ENDIAN(28) + stxvw4x(60, 0, %[y]) + +: [buf1] "+b" (buf1), [buf2] "+b" (buf2) +: [y] "b" (y), [h] "b" (h), [num4] "b" (num4), [num1] "b" (num1), + [cc0] "b" (cc0), [cc1] "b" (cc1), [cc2] "b" (cc2), [cc3] "b" (cc3) +#if BR_POWER8_LE + , [idx2be] "b" (idx2be) +#endif +: "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8", "v9", + "v10", "v11", "v12", "v13", "v14", "v15", "v16", "v17", "v18", "v19", + "v20", "v21", "v22", "v23", "v24", "v25", "v26", "v27", "v28", "v29", + "ctr", "memory" + ); +} + +/* see bearssl_hash.h */ +br_ghash +br_ghash_pwr8_get(void) +{ + return &br_ghash_pwr8; +} + +#else + +/* see bearssl_hash.h */ +br_ghash +br_ghash_pwr8_get(void) +{ + return 0; +} + +#endif diff --git a/test/monniaux/BearSSL/src/hash/md5.c b/test/monniaux/BearSSL/src/hash/md5.c new file mode 100644 index 00000000..0df7abe0 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/md5.c @@ -0,0 +1,208 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +#define F(B, C, D) ((((C) ^ (D)) & (B)) ^ (D)) +#define G(B, C, D) ((((C) ^ (B)) & (D)) ^ (C)) +#define H(B, C, D) ((B) ^ (C) ^ (D)) +#define I(B, C, D) ((C) ^ ((B) | ~(D))) + +#define ROTL(x, n) (((x) << (n)) | ((x) >> (32 - (n)))) + +/* see inner.h */ +const uint32_t br_md5_IV[4] = { + 0x67452301, 0xEFCDAB89, 0x98BADCFE, 0x10325476 +}; + +static const uint32_t K[64] = { + 0xD76AA478, 0xE8C7B756, 0x242070DB, 0xC1BDCEEE, + 0xF57C0FAF, 0x4787C62A, 0xA8304613, 0xFD469501, + 0x698098D8, 0x8B44F7AF, 0xFFFF5BB1, 0x895CD7BE, + 0x6B901122, 0xFD987193, 0xA679438E, 0x49B40821, + + 0xF61E2562, 0xC040B340, 0x265E5A51, 0xE9B6C7AA, + 0xD62F105D, 0x02441453, 0xD8A1E681, 0xE7D3FBC8, + 0x21E1CDE6, 0xC33707D6, 0xF4D50D87, 0x455A14ED, + 0xA9E3E905, 0xFCEFA3F8, 0x676F02D9, 0x8D2A4C8A, + + 0xFFFA3942, 0x8771F681, 0x6D9D6122, 0xFDE5380C, + 0xA4BEEA44, 0x4BDECFA9, 0xF6BB4B60, 0xBEBFBC70, + 0x289B7EC6, 0xEAA127FA, 0xD4EF3085, 0x04881D05, + 0xD9D4D039, 0xE6DB99E5, 0x1FA27CF8, 0xC4AC5665, + + 0xF4292244, 0x432AFF97, 0xAB9423A7, 0xFC93A039, + 0x655B59C3, 0x8F0CCC92, 0xFFEFF47D, 0x85845DD1, + 0x6FA87E4F, 0xFE2CE6E0, 0xA3014314, 0x4E0811A1, + 0xF7537E82, 0xBD3AF235, 0x2AD7D2BB, 0xEB86D391 +}; + +static const unsigned char MP[48] = { + 1, 6, 11, 0, 5, 10, 15, 4, 9, 14, 3, 8, 13, 2, 7, 12, + 5, 8, 11, 14, 1, 4, 7, 10, 13, 0, 3, 6, 9, 12, 15, 2, + 0, 7, 14, 5, 12, 3, 10, 1, 8, 15, 6, 13, 4, 11, 2, 9 +}; + +/* see inner.h */ +void +br_md5_round(const unsigned char *buf, uint32_t *val) +{ + uint32_t m[16]; + uint32_t a, b, c, d; + int i; + + a = val[0]; + b = val[1]; + c = val[2]; + d = val[3]; + /* obsolete + for (i = 0; i < 16; i ++) { + m[i] = br_dec32le(buf + (i << 2)); + } + */ + br_range_dec32le(m, 16, buf); + + for (i = 0; i < 16; i += 4) { + a = b + ROTL(a + F(b, c, d) + m[i + 0] + K[i + 0], 7); + d = a + ROTL(d + F(a, b, c) + m[i + 1] + K[i + 1], 12); + c = d + ROTL(c + F(d, a, b) + m[i + 2] + K[i + 2], 17); + b = c + ROTL(b + F(c, d, a) + m[i + 3] + K[i + 3], 22); + } + for (i = 16; i < 32; i += 4) { + a = b + ROTL(a + G(b, c, d) + m[MP[i - 16]] + K[i + 0], 5); + d = a + ROTL(d + G(a, b, c) + m[MP[i - 15]] + K[i + 1], 9); + c = d + ROTL(c + G(d, a, b) + m[MP[i - 14]] + K[i + 2], 14); + b = c + ROTL(b + G(c, d, a) + m[MP[i - 13]] + K[i + 3], 20); + } + for (i = 32; i < 48; i += 4) { + a = b + ROTL(a + H(b, c, d) + m[MP[i - 16]] + K[i + 0], 4); + d = a + ROTL(d + H(a, b, c) + m[MP[i - 15]] + K[i + 1], 11); + c = d + ROTL(c + H(d, a, b) + m[MP[i - 14]] + K[i + 2], 16); + b = c + ROTL(b + H(c, d, a) + m[MP[i - 13]] + K[i + 3], 23); + } + for (i = 48; i < 64; i += 4) { + a = b + ROTL(a + I(b, c, d) + m[MP[i - 16]] + K[i + 0], 6); + d = a + ROTL(d + I(a, b, c) + m[MP[i - 15]] + K[i + 1], 10); + c = d + ROTL(c + I(d, a, b) + m[MP[i - 14]] + K[i + 2], 15); + b = c + ROTL(b + I(c, d, a) + m[MP[i - 13]] + K[i + 3], 21); + } + + val[0] += a; + val[1] += b; + val[2] += c; + val[3] += d; +} + +/* see bearssl.h */ +void +br_md5_init(br_md5_context *cc) +{ + cc->vtable = &br_md5_vtable; + memcpy(cc->val, br_md5_IV, sizeof cc->val); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_md5_update(br_md5_context *cc, const void *data, size_t len) +{ + const unsigned char *buf; + size_t ptr; + + buf = data; + ptr = (size_t)cc->count & 63; + while (len > 0) { + size_t clen; + + clen = 64 - ptr; + if (clen > len) { + clen = len; + } + memcpy(cc->buf + ptr, buf, clen); + ptr += clen; + buf += clen; + len -= clen; + cc->count += (uint64_t)clen; + if (ptr == 64) { + br_md5_round(cc->buf, cc->val); + ptr = 0; + } + } +} + +/* see bearssl.h */ +void +br_md5_out(const br_md5_context *cc, void *dst) +{ + unsigned char buf[64]; + uint32_t val[4]; + size_t ptr; + + ptr = (size_t)cc->count & 63; + memcpy(buf, cc->buf, ptr); + memcpy(val, cc->val, sizeof val); + buf[ptr ++] = 0x80; + if (ptr > 56) { + memset(buf + ptr, 0, 64 - ptr); + br_md5_round(buf, val); + memset(buf, 0, 56); + } else { + memset(buf + ptr, 0, 56 - ptr); + } + br_enc64le(buf + 56, cc->count << 3); + br_md5_round(buf, val); + br_range_enc32le(dst, val, 4); +} + +/* see bearssl.h */ +uint64_t +br_md5_state(const br_md5_context *cc, void *dst) +{ + br_range_enc32le(dst, cc->val, 4); + return cc->count; +} + +/* see bearssl.h */ +void +br_md5_set_state(br_md5_context *cc, const void *stb, uint64_t count) +{ + br_range_dec32le(cc->val, 4, stb); + cc->count = count; +} + +/* see bearssl.h */ +const br_hash_class br_md5_vtable = { + sizeof(br_md5_context), + BR_HASHDESC_ID(br_md5_ID) + | BR_HASHDESC_OUT(16) + | BR_HASHDESC_STATE(16) + | BR_HASHDESC_LBLEN(6) + | BR_HASHDESC_MD_PADDING, + (void (*)(const br_hash_class **))&br_md5_init, + (void (*)(const br_hash_class **, const void *, size_t))&br_md5_update, + (void (*)(const br_hash_class *const *, void *))&br_md5_out, + (uint64_t (*)(const br_hash_class *const *, void *))&br_md5_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_md5_set_state +}; diff --git a/test/monniaux/BearSSL/src/hash/md5sha1.c b/test/monniaux/BearSSL/src/hash/md5sha1.c new file mode 100644 index 00000000..f701aeed --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/md5sha1.c @@ -0,0 +1,141 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* see bearssl.h */ +void +br_md5sha1_init(br_md5sha1_context *cc) +{ + cc->vtable = &br_md5sha1_vtable; + memcpy(cc->val_md5, br_md5_IV, sizeof cc->val_md5); + memcpy(cc->val_sha1, br_sha1_IV, sizeof cc->val_sha1); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_md5sha1_update(br_md5sha1_context *cc, const void *data, size_t len) +{ + const unsigned char *buf; + size_t ptr; + + buf = data; + ptr = (size_t)cc->count & 63; + while (len > 0) { + size_t clen; + + clen = 64 - ptr; + if (clen > len) { + clen = len; + } + memcpy(cc->buf + ptr, buf, clen); + ptr += clen; + buf += clen; + len -= clen; + cc->count += (uint64_t)clen; + if (ptr == 64) { + br_md5_round(cc->buf, cc->val_md5); + br_sha1_round(cc->buf, cc->val_sha1); + ptr = 0; + } + } +} + +/* see bearssl.h */ +void +br_md5sha1_out(const br_md5sha1_context *cc, void *dst) +{ + unsigned char buf[64]; + uint32_t val_md5[4]; + uint32_t val_sha1[5]; + size_t ptr; + unsigned char *out; + uint64_t count; + + count = cc->count; + ptr = (size_t)count & 63; + memcpy(buf, cc->buf, ptr); + memcpy(val_md5, cc->val_md5, sizeof val_md5); + memcpy(val_sha1, cc->val_sha1, sizeof val_sha1); + buf[ptr ++] = 0x80; + if (ptr > 56) { + memset(buf + ptr, 0, 64 - ptr); + br_md5_round(buf, val_md5); + br_sha1_round(buf, val_sha1); + memset(buf, 0, 56); + } else { + memset(buf + ptr, 0, 56 - ptr); + } + count <<= 3; + br_enc64le(buf + 56, count); + br_md5_round(buf, val_md5); + br_enc64be(buf + 56, count); + br_sha1_round(buf, val_sha1); + out = dst; + br_range_enc32le(out, val_md5, 4); + br_range_enc32be(out + 16, val_sha1, 5); +} + +/* see bearssl.h */ +uint64_t +br_md5sha1_state(const br_md5sha1_context *cc, void *dst) +{ + unsigned char *out; + + out = dst; + br_range_enc32le(out, cc->val_md5, 4); + br_range_enc32be(out + 16, cc->val_sha1, 5); + return cc->count; +} + +/* see bearssl.h */ +void +br_md5sha1_set_state(br_md5sha1_context *cc, const void *stb, uint64_t count) +{ + const unsigned char *buf; + + buf = stb; + br_range_dec32le(cc->val_md5, 4, buf); + br_range_dec32be(cc->val_sha1, 5, buf + 16); + cc->count = count; +} + +/* see bearssl.h */ +const br_hash_class br_md5sha1_vtable = { + sizeof(br_md5sha1_context), + BR_HASHDESC_ID(br_md5sha1_ID) + | BR_HASHDESC_OUT(36) + | BR_HASHDESC_STATE(36) + | BR_HASHDESC_LBLEN(6), + (void (*)(const br_hash_class **))&br_md5sha1_init, + (void (*)(const br_hash_class **, const void *, size_t)) + &br_md5sha1_update, + (void (*)(const br_hash_class *const *, void *)) + &br_md5sha1_out, + (uint64_t (*)(const br_hash_class *const *, void *)) + &br_md5sha1_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_md5sha1_set_state +}; diff --git a/test/monniaux/BearSSL/src/hash/mgf1.c b/test/monniaux/BearSSL/src/hash/mgf1.c new file mode 100644 index 00000000..7a235887 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/mgf1.c @@ -0,0 +1,56 @@ +/* + * Copyright (c) 2018 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* see inner.h */ +void +br_mgf1_xor(void *data, size_t len, + const br_hash_class *dig, const void *seed, size_t seed_len) +{ + unsigned char *buf; + size_t u, hlen; + uint32_t c; + + buf = data; + hlen = br_digest_size(dig); + for (u = 0, c = 0; u < len; u += hlen, c ++) { + br_hash_compat_context hc; + unsigned char tmp[64]; + size_t v; + + hc.vtable = dig; + dig->init(&hc.vtable); + dig->update(&hc.vtable, seed, seed_len); + br_enc32be(tmp, c); + dig->update(&hc.vtable, tmp, 4); + dig->out(&hc.vtable, tmp); + for (v = 0; v < hlen; v ++) { + if ((u + v) >= len) { + break; + } + buf[u + v] ^= tmp[v]; + } + } +} diff --git a/test/monniaux/BearSSL/src/hash/multihash.c b/test/monniaux/BearSSL/src/hash/multihash.c new file mode 100644 index 00000000..b6df2e0e --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/multihash.c @@ -0,0 +1,166 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +/* + * An aggregate context that is large enough for all supported hash + * functions. + */ +typedef union { + const br_hash_class *vtable; + br_md5_context md5; + br_sha1_context sha1; + br_sha224_context sha224; + br_sha256_context sha256; + br_sha384_context sha384; + br_sha512_context sha512; +} gen_hash_context; + +/* + * Get the offset to the state for a specific hash function within the + * context structure. This shall be called only for the supported hash + * functions, + */ +static size_t +get_state_offset(int id) +{ + if (id >= 5) { + /* + * SHA-384 has id 5, and SHA-512 has id 6. Both use + * eight 64-bit words for their state. + */ + return offsetof(br_multihash_context, val_64) + + ((size_t)(id - 5) * (8 * sizeof(uint64_t))); + } else { + /* + * MD5 has id 1, SHA-1 has id 2, SHA-224 has id 3 and + * SHA-256 has id 4. They use 32-bit words for their + * states (4 words for MD5, 5 for SHA-1, 8 for SHA-224 + * and 8 for SHA-256). + */ + unsigned x; + + x = id - 1; + x = ((x + (x & (x >> 1))) << 2) + (x >> 1); + return offsetof(br_multihash_context, val_32) + + x * sizeof(uint32_t); + } +} + +/* see bearssl_hash.h */ +void +br_multihash_zero(br_multihash_context *ctx) +{ + /* + * This is not standard, but yields very short and efficient code, + * and it works "everywhere". + */ + memset(ctx, 0, sizeof *ctx); +} + +/* see bearssl_hash.h */ +void +br_multihash_init(br_multihash_context *ctx) +{ + int i; + + ctx->count = 0; + for (i = 1; i <= 6; i ++) { + const br_hash_class *hc; + + hc = ctx->impl[i - 1]; + if (hc != NULL) { + gen_hash_context g; + + hc->init(&g.vtable); + hc->state(&g.vtable, + (unsigned char *)ctx + get_state_offset(i)); + } + } +} + +/* see bearssl_hash.h */ +void +br_multihash_update(br_multihash_context *ctx, const void *data, size_t len) +{ + const unsigned char *buf; + size_t ptr; + + buf = data; + ptr = (size_t)ctx->count & 127; + while (len > 0) { + size_t clen; + + clen = 128 - ptr; + if (clen > len) { + clen = len; + } + memcpy(ctx->buf + ptr, buf, clen); + ptr += clen; + buf += clen; + len -= clen; + ctx->count += (uint64_t)clen; + if (ptr == 128) { + int i; + + for (i = 1; i <= 6; i ++) { + const br_hash_class *hc; + + hc = ctx->impl[i - 1]; + if (hc != NULL) { + gen_hash_context g; + unsigned char *state; + + state = (unsigned char *)ctx + + get_state_offset(i); + hc->set_state(&g.vtable, + state, ctx->count - 128); + hc->update(&g.vtable, ctx->buf, 128); + hc->state(&g.vtable, state); + } + } + ptr = 0; + } + } +} + +/* see bearssl_hash.h */ +size_t +br_multihash_out(const br_multihash_context *ctx, int id, void *dst) +{ + const br_hash_class *hc; + gen_hash_context g; + const unsigned char *state; + + hc = ctx->impl[id - 1]; + if (hc == NULL) { + return 0; + } + state = (const unsigned char *)ctx + get_state_offset(id); + hc->set_state(&g.vtable, state, ctx->count & ~(uint64_t)127); + hc->update(&g.vtable, ctx->buf, ctx->count & (uint64_t)127); + hc->out(&g.vtable, dst); + return (hc->desc >> BR_HASHDESC_OUT_OFF) & BR_HASHDESC_OUT_MASK; +} diff --git a/test/monniaux/BearSSL/src/hash/sha1.c b/test/monniaux/BearSSL/src/hash/sha1.c new file mode 100644 index 00000000..4f65d846 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/sha1.c @@ -0,0 +1,191 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +#define F(B, C, D) ((((C) ^ (D)) & (B)) ^ (D)) +#define G(B, C, D) ((B) ^ (C) ^ (D)) +#define H(B, C, D) (((D) & (C)) | (((D) | (C)) & (B))) +#define I(B, C, D) G(B, C, D) + +#define ROTL(x, n) (((x) << (n)) | ((x) >> (32 - (n)))) + +#define K1 ((uint32_t)0x5A827999) +#define K2 ((uint32_t)0x6ED9EBA1) +#define K3 ((uint32_t)0x8F1BBCDC) +#define K4 ((uint32_t)0xCA62C1D6) + +/* see inner.h */ +const uint32_t br_sha1_IV[5] = { + 0x67452301, 0xEFCDAB89, 0x98BADCFE, 0x10325476, 0xC3D2E1F0 +}; + +/* see inner.h */ +void +br_sha1_round(const unsigned char *buf, uint32_t *val) +{ + uint32_t m[80]; + uint32_t a, b, c, d, e; + int i; + + a = val[0]; + b = val[1]; + c = val[2]; + d = val[3]; + e = val[4]; + br_range_dec32be(m, 16, buf); + for (i = 16; i < 80; i ++) { + uint32_t x = m[i - 3] ^ m[i - 8] ^ m[i - 14] ^ m[i - 16]; + m[i] = ROTL(x, 1); + } + + for (i = 0; i < 20; i += 5) { + e += ROTL(a, 5) + F(b, c, d) + K1 + m[i + 0]; b = ROTL(b, 30); + d += ROTL(e, 5) + F(a, b, c) + K1 + m[i + 1]; a = ROTL(a, 30); + c += ROTL(d, 5) + F(e, a, b) + K1 + m[i + 2]; e = ROTL(e, 30); + b += ROTL(c, 5) + F(d, e, a) + K1 + m[i + 3]; d = ROTL(d, 30); + a += ROTL(b, 5) + F(c, d, e) + K1 + m[i + 4]; c = ROTL(c, 30); + } + for (i = 20; i < 40; i += 5) { + e += ROTL(a, 5) + G(b, c, d) + K2 + m[i + 0]; b = ROTL(b, 30); + d += ROTL(e, 5) + G(a, b, c) + K2 + m[i + 1]; a = ROTL(a, 30); + c += ROTL(d, 5) + G(e, a, b) + K2 + m[i + 2]; e = ROTL(e, 30); + b += ROTL(c, 5) + G(d, e, a) + K2 + m[i + 3]; d = ROTL(d, 30); + a += ROTL(b, 5) + G(c, d, e) + K2 + m[i + 4]; c = ROTL(c, 30); + } + for (i = 40; i < 60; i += 5) { + e += ROTL(a, 5) + H(b, c, d) + K3 + m[i + 0]; b = ROTL(b, 30); + d += ROTL(e, 5) + H(a, b, c) + K3 + m[i + 1]; a = ROTL(a, 30); + c += ROTL(d, 5) + H(e, a, b) + K3 + m[i + 2]; e = ROTL(e, 30); + b += ROTL(c, 5) + H(d, e, a) + K3 + m[i + 3]; d = ROTL(d, 30); + a += ROTL(b, 5) + H(c, d, e) + K3 + m[i + 4]; c = ROTL(c, 30); + } + for (i = 60; i < 80; i += 5) { + e += ROTL(a, 5) + I(b, c, d) + K4 + m[i + 0]; b = ROTL(b, 30); + d += ROTL(e, 5) + I(a, b, c) + K4 + m[i + 1]; a = ROTL(a, 30); + c += ROTL(d, 5) + I(e, a, b) + K4 + m[i + 2]; e = ROTL(e, 30); + b += ROTL(c, 5) + I(d, e, a) + K4 + m[i + 3]; d = ROTL(d, 30); + a += ROTL(b, 5) + I(c, d, e) + K4 + m[i + 4]; c = ROTL(c, 30); + } + + val[0] += a; + val[1] += b; + val[2] += c; + val[3] += d; + val[4] += e; +} + +/* see bearssl.h */ +void +br_sha1_init(br_sha1_context *cc) +{ + cc->vtable = &br_sha1_vtable; + memcpy(cc->val, br_sha1_IV, sizeof cc->val); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_sha1_update(br_sha1_context *cc, const void *data, size_t len) +{ + const unsigned char *buf; + size_t ptr; + + buf = data; + ptr = (size_t)cc->count & 63; + while (len > 0) { + size_t clen; + + clen = 64 - ptr; + if (clen > len) { + clen = len; + } + memcpy(cc->buf + ptr, buf, clen); + ptr += clen; + buf += clen; + len -= clen; + cc->count += (uint64_t)clen; + if (ptr == 64) { + br_sha1_round(cc->buf, cc->val); + ptr = 0; + } + } +} + +/* see bearssl.h */ +void +br_sha1_out(const br_sha1_context *cc, void *dst) +{ + unsigned char buf[64]; + uint32_t val[5]; + size_t ptr; + + ptr = (size_t)cc->count & 63; + memcpy(buf, cc->buf, ptr); + memcpy(val, cc->val, sizeof val); + buf[ptr ++] = 0x80; + if (ptr > 56) { + memset(buf + ptr, 0, 64 - ptr); + br_sha1_round(buf, val); + memset(buf, 0, 56); + } else { + memset(buf + ptr, 0, 56 - ptr); + } + br_enc64be(buf + 56, cc->count << 3); + br_sha1_round(buf, val); + br_range_enc32be(dst, val, 5); +} + +/* see bearssl.h */ +uint64_t +br_sha1_state(const br_sha1_context *cc, void *dst) +{ + br_range_enc32be(dst, cc->val, 5); + return cc->count; +} + +/* see bearssl.h */ +void +br_sha1_set_state(br_sha1_context *cc, const void *stb, uint64_t count) +{ + br_range_dec32be(cc->val, 5, stb); + cc->count = count; +} + +/* see bearssl.h */ +const br_hash_class br_sha1_vtable = { + sizeof(br_sha1_context), + BR_HASHDESC_ID(br_sha1_ID) + | BR_HASHDESC_OUT(20) + | BR_HASHDESC_STATE(20) + | BR_HASHDESC_LBLEN(6) + | BR_HASHDESC_MD_PADDING + | BR_HASHDESC_MD_PADDING_BE, + (void (*)(const br_hash_class **))&br_sha1_init, + (void (*)(const br_hash_class **, const void *, size_t))&br_sha1_update, + (void (*)(const br_hash_class *const *, void *))&br_sha1_out, + (uint64_t (*)(const br_hash_class *const *, void *))&br_sha1_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_sha1_set_state +}; diff --git a/test/monniaux/BearSSL/src/hash/sha2big.c b/test/monniaux/BearSSL/src/hash/sha2big.c new file mode 100644 index 00000000..5be92ed5 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/sha2big.c @@ -0,0 +1,285 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +#define CH(X, Y, Z) ((((Y) ^ (Z)) & (X)) ^ (Z)) +#define MAJ(X, Y, Z) (((Y) & (Z)) | (((Y) | (Z)) & (X))) + +#define ROTR(x, n) (((uint64_t)(x) << (64 - (n))) | ((uint64_t)(x) >> (n))) + +#define BSG5_0(x) (ROTR(x, 28) ^ ROTR(x, 34) ^ ROTR(x, 39)) +#define BSG5_1(x) (ROTR(x, 14) ^ ROTR(x, 18) ^ ROTR(x, 41)) +#define SSG5_0(x) (ROTR(x, 1) ^ ROTR(x, 8) ^ (uint64_t)((x) >> 7)) +#define SSG5_1(x) (ROTR(x, 19) ^ ROTR(x, 61) ^ (uint64_t)((x) >> 6)) + +static const uint64_t IV384[8] = { + 0xCBBB9D5DC1059ED8, 0x629A292A367CD507, + 0x9159015A3070DD17, 0x152FECD8F70E5939, + 0x67332667FFC00B31, 0x8EB44A8768581511, + 0xDB0C2E0D64F98FA7, 0x47B5481DBEFA4FA4 +}; + +static const uint64_t IV512[8] = { + 0x6A09E667F3BCC908, 0xBB67AE8584CAA73B, + 0x3C6EF372FE94F82B, 0xA54FF53A5F1D36F1, + 0x510E527FADE682D1, 0x9B05688C2B3E6C1F, + 0x1F83D9ABFB41BD6B, 0x5BE0CD19137E2179 +}; + +static const uint64_t K[80] = { + 0x428A2F98D728AE22, 0x7137449123EF65CD, + 0xB5C0FBCFEC4D3B2F, 0xE9B5DBA58189DBBC, + 0x3956C25BF348B538, 0x59F111F1B605D019, + 0x923F82A4AF194F9B, 0xAB1C5ED5DA6D8118, + 0xD807AA98A3030242, 0x12835B0145706FBE, + 0x243185BE4EE4B28C, 0x550C7DC3D5FFB4E2, + 0x72BE5D74F27B896F, 0x80DEB1FE3B1696B1, + 0x9BDC06A725C71235, 0xC19BF174CF692694, + 0xE49B69C19EF14AD2, 0xEFBE4786384F25E3, + 0x0FC19DC68B8CD5B5, 0x240CA1CC77AC9C65, + 0x2DE92C6F592B0275, 0x4A7484AA6EA6E483, + 0x5CB0A9DCBD41FBD4, 0x76F988DA831153B5, + 0x983E5152EE66DFAB, 0xA831C66D2DB43210, + 0xB00327C898FB213F, 0xBF597FC7BEEF0EE4, + 0xC6E00BF33DA88FC2, 0xD5A79147930AA725, + 0x06CA6351E003826F, 0x142929670A0E6E70, + 0x27B70A8546D22FFC, 0x2E1B21385C26C926, + 0x4D2C6DFC5AC42AED, 0x53380D139D95B3DF, + 0x650A73548BAF63DE, 0x766A0ABB3C77B2A8, + 0x81C2C92E47EDAEE6, 0x92722C851482353B, + 0xA2BFE8A14CF10364, 0xA81A664BBC423001, + 0xC24B8B70D0F89791, 0xC76C51A30654BE30, + 0xD192E819D6EF5218, 0xD69906245565A910, + 0xF40E35855771202A, 0x106AA07032BBD1B8, + 0x19A4C116B8D2D0C8, 0x1E376C085141AB53, + 0x2748774CDF8EEB99, 0x34B0BCB5E19B48A8, + 0x391C0CB3C5C95A63, 0x4ED8AA4AE3418ACB, + 0x5B9CCA4F7763E373, 0x682E6FF3D6B2B8A3, + 0x748F82EE5DEFB2FC, 0x78A5636F43172F60, + 0x84C87814A1F0AB72, 0x8CC702081A6439EC, + 0x90BEFFFA23631E28, 0xA4506CEBDE82BDE9, + 0xBEF9A3F7B2C67915, 0xC67178F2E372532B, + 0xCA273ECEEA26619C, 0xD186B8C721C0C207, + 0xEADA7DD6CDE0EB1E, 0xF57D4F7FEE6ED178, + 0x06F067AA72176FBA, 0x0A637DC5A2C898A6, + 0x113F9804BEF90DAE, 0x1B710B35131C471B, + 0x28DB77F523047D84, 0x32CAAB7B40C72493, + 0x3C9EBE0A15C9BEBC, 0x431D67C49C100D4C, + 0x4CC5D4BECB3E42B6, 0x597F299CFC657E2A, + 0x5FCB6FAB3AD6FAEC, 0x6C44198C4A475817 +}; + +static void +sha2big_round(const unsigned char *buf, uint64_t *val) +{ + +#define SHA2BIG_STEP(A, B, C, D, E, F, G, H, j) do { \ + uint64_t T1, T2; \ + T1 = H + BSG5_1(E) + CH(E, F, G) + K[j] + w[j]; \ + T2 = BSG5_0(A) + MAJ(A, B, C); \ + D += T1; \ + H = T1 + T2; \ + } while (0) + + int i; + uint64_t a, b, c, d, e, f, g, h; + uint64_t w[80]; + + br_range_dec64be(w, 16, buf); + for (i = 16; i < 80; i ++) { + w[i] = SSG5_1(w[i - 2]) + w[i - 7] + + SSG5_0(w[i - 15]) + w[i - 16]; + } + a = val[0]; + b = val[1]; + c = val[2]; + d = val[3]; + e = val[4]; + f = val[5]; + g = val[6]; + h = val[7]; + for (i = 0; i < 80; i += 8) { + SHA2BIG_STEP(a, b, c, d, e, f, g, h, i + 0); + SHA2BIG_STEP(h, a, b, c, d, e, f, g, i + 1); + SHA2BIG_STEP(g, h, a, b, c, d, e, f, i + 2); + SHA2BIG_STEP(f, g, h, a, b, c, d, e, i + 3); + SHA2BIG_STEP(e, f, g, h, a, b, c, d, i + 4); + SHA2BIG_STEP(d, e, f, g, h, a, b, c, i + 5); + SHA2BIG_STEP(c, d, e, f, g, h, a, b, i + 6); + SHA2BIG_STEP(b, c, d, e, f, g, h, a, i + 7); + } + val[0] += a; + val[1] += b; + val[2] += c; + val[3] += d; + val[4] += e; + val[5] += f; + val[6] += g; + val[7] += h; +} + +static void +sha2big_update(br_sha384_context *cc, const void *data, size_t len) +{ + const unsigned char *buf; + size_t ptr; + + buf = data; + ptr = (size_t)cc->count & 127; + cc->count += (uint64_t)len; + while (len > 0) { + size_t clen; + + clen = 128 - ptr; + if (clen > len) { + clen = len; + } + memcpy(cc->buf + ptr, buf, clen); + ptr += clen; + buf += clen; + len -= clen; + if (ptr == 128) { + sha2big_round(cc->buf, cc->val); + ptr = 0; + } + } +} + +static void +sha2big_out(const br_sha384_context *cc, void *dst, int num) +{ + unsigned char buf[128]; + uint64_t val[8]; + size_t ptr; + + ptr = (size_t)cc->count & 127; + memcpy(buf, cc->buf, ptr); + memcpy(val, cc->val, sizeof val); + buf[ptr ++] = 0x80; + if (ptr > 112) { + memset(buf + ptr, 0, 128 - ptr); + sha2big_round(buf, val); + memset(buf, 0, 112); + } else { + memset(buf + ptr, 0, 112 - ptr); + } + br_enc64be(buf + 112, cc->count >> 61); + br_enc64be(buf + 120, cc->count << 3); + sha2big_round(buf, val); + br_range_enc64be(dst, val, num); +} + +/* see bearssl.h */ +void +br_sha384_init(br_sha384_context *cc) +{ + cc->vtable = &br_sha384_vtable; + memcpy(cc->val, IV384, sizeof IV384); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_sha384_update(br_sha384_context *cc, const void *data, size_t len) +{ + sha2big_update(cc, data, len); +} + +/* see bearssl.h */ +void +br_sha384_out(const br_sha384_context *cc, void *dst) +{ + sha2big_out(cc, dst, 6); +} + +/* see bearssl.h */ +uint64_t +br_sha384_state(const br_sha384_context *cc, void *dst) +{ + br_range_enc64be(dst, cc->val, 8); + return cc->count; +} + +/* see bearssl.h */ +void +br_sha384_set_state(br_sha384_context *cc, const void *stb, uint64_t count) +{ + br_range_dec64be(cc->val, 8, stb); + cc->count = count; +} + +/* see bearssl.h */ +void +br_sha512_init(br_sha512_context *cc) +{ + cc->vtable = &br_sha512_vtable; + memcpy(cc->val, IV512, sizeof IV512); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_sha512_out(const br_sha512_context *cc, void *dst) +{ + sha2big_out(cc, dst, 8); +} + +/* see bearssl.h */ +const br_hash_class br_sha384_vtable = { + sizeof(br_sha384_context), + BR_HASHDESC_ID(br_sha384_ID) + | BR_HASHDESC_OUT(48) + | BR_HASHDESC_STATE(64) + | BR_HASHDESC_LBLEN(7) + | BR_HASHDESC_MD_PADDING + | BR_HASHDESC_MD_PADDING_BE + | BR_HASHDESC_MD_PADDING_128, + (void (*)(const br_hash_class **))&br_sha384_init, + (void (*)(const br_hash_class **, const void *, size_t)) + &br_sha384_update, + (void (*)(const br_hash_class *const *, void *))&br_sha384_out, + (uint64_t (*)(const br_hash_class *const *, void *))&br_sha384_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_sha384_set_state +}; + +/* see bearssl.h */ +const br_hash_class br_sha512_vtable = { + sizeof(br_sha512_context), + BR_HASHDESC_ID(br_sha512_ID) + | BR_HASHDESC_OUT(64) + | BR_HASHDESC_STATE(64) + | BR_HASHDESC_LBLEN(7) + | BR_HASHDESC_MD_PADDING + | BR_HASHDESC_MD_PADDING_BE + | BR_HASHDESC_MD_PADDING_128, + (void (*)(const br_hash_class **))&br_sha512_init, + (void (*)(const br_hash_class **, const void *, size_t)) + &br_sha512_update, + (void (*)(const br_hash_class *const *, void *))&br_sha512_out, + (uint64_t (*)(const br_hash_class *const *, void *))&br_sha512_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_sha512_set_state +}; diff --git a/test/monniaux/BearSSL/src/hash/sha2small.c b/test/monniaux/BearSSL/src/hash/sha2small.c new file mode 100644 index 00000000..ca196559 --- /dev/null +++ b/test/monniaux/BearSSL/src/hash/sha2small.c @@ -0,0 +1,341 @@ +/* + * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> + * + * 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 "inner.h" + +#define CH(X, Y, Z) ((((Y) ^ (Z)) & (X)) ^ (Z)) +#define MAJ(X, Y, Z) (((Y) & (Z)) | (((Y) | (Z)) & (X))) + +#define ROTR(x, n) (((uint32_t)(x) << (32 - (n))) | ((uint32_t)(x) >> (n))) + +#define BSG2_0(x) (ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22)) +#define BSG2_1(x) (ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25)) +#define SSG2_0(x) (ROTR(x, 7) ^ ROTR(x, 18) ^ (uint32_t)((x) >> 3)) +#define SSG2_1(x) (ROTR(x, 17) ^ ROTR(x, 19) ^ (uint32_t)((x) >> 10)) + +/* see inner.h */ +const uint32_t br_sha224_IV[8] = { + 0xC1059ED8, 0x367CD507, 0x3070DD17, 0xF70E5939, + 0xFFC00B31, 0x68581511, 0x64F98FA7, 0xBEFA4FA4 +}; + +/* see inner.h */ +const uint32_t br_sha256_IV[8] = { + 0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A, + 0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19 +}; + +static const uint32_t K[64] = { + 0x428A2F98, 0x71374491, 0xB5C0FBCF, 0xE9B5DBA5, + 0x3956C25B, 0x59F111F1, 0x923F82A4, 0xAB1C5ED5, + 0xD807AA98, 0x12835B01, 0x243185BE, 0x550C7DC3, + 0x72BE5D74, 0x80DEB1FE, 0x9BDC06A7, 0xC19BF174, + 0xE49B69C1, 0xEFBE4786, 0x0FC19DC6, 0x240CA1CC, + 0x2DE92C6F, 0x4A7484AA, 0x5CB0A9DC, 0x76F988DA, + 0x983E5152, 0xA831C66D, 0xB00327C8, 0xBF597FC7, + 0xC6E00BF3, 0xD5A79147, 0x06CA6351, 0x14292967, + 0x27B70A85, 0x2E1B2138, 0x4D2C6DFC, 0x53380D13, + 0x650A7354, 0x766A0ABB, 0x81C2C92E, 0x92722C85, + 0xA2BFE8A1, 0xA81A664B, 0xC24B8B70, 0xC76C51A3, + 0xD192E819, 0xD6990624, 0xF40E3585, 0x106AA070, + 0x19A4C116, 0x1E376C08, 0x2748774C, 0x34B0BCB5, + 0x391C0CB3, 0x4ED8AA4A, 0x5B9CCA4F, 0x682E6FF3, + 0x748F82EE, 0x78A5636F, 0x84C87814, 0x8CC70208, + 0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2 +}; + +/* see inner.h */ +void +br_sha2small_round(const unsigned char *buf, uint32_t *val) +{ + +#define SHA2_STEP(A, B, C, D, E, F, G, H, j) do { \ + uint32_t T1, T2; \ + T1 = H + BSG2_1(E) + CH(E, F, G) + K[j] + w[j]; \ + T2 = BSG2_0(A) + MAJ(A, B, C); \ + D += T1; \ + H = T1 + T2; \ + } while (0) + + int i; + uint32_t a, b, c, d, e, f, g, h; + uint32_t w[64]; + + br_range_dec32be(w, 16, buf); + for (i = 16; i < 64; i ++) { + w[i] = SSG2_1(w[i - 2]) + w[i - 7] + + SSG2_0(w[i - 15]) + w[i - 16]; + } + a = val[0]; + b = val[1]; + c = val[2]; + d = val[3]; + e = val[4]; + f = val[5]; + g = val[6]; + h = val[7]; + for (i = 0; i < 64; i += 8) { + SHA2_STEP(a, b, c, d, e, f, g, h, i + 0); + SHA2_STEP(h, a, b, c, d, e, f, g, i + 1); + SHA2_STEP(g, h, a, b, c, d, e, f, i + 2); + SHA2_STEP(f, g, h, a, b, c, d, e, i + 3); + SHA2_STEP(e, f, g, h, a, b, c, d, i + 4); + SHA2_STEP(d, e, f, g, h, a, b, c, i + 5); + SHA2_STEP(c, d, e, f, g, h, a, b, i + 6); + SHA2_STEP(b, c, d, e, f, g, h, a, i + 7); + } + val[0] += a; + val[1] += b; + val[2] += c; + val[3] += d; + val[4] += e; + val[5] += f; + val[6] += g; + val[7] += h; + +#if 0 +/* obsolete */ +#define SHA2_MEXP1(pc) do { \ + W[pc] = br_dec32be(buf + ((pc) << 2)); \ + } while (0) + +#define SHA2_MEXP2(pc) do { \ + W[(pc) & 0x0F] = SSG2_1(W[((pc) - 2) & 0x0F]) \ + + W[((pc) - 7) & 0x0F] \ + + SSG2_0(W[((pc) - 15) & 0x0F]) + W[(pc) & 0x0F]; \ + } while (0) + +#define SHA2_STEPn(n, a, b, c, d, e, f, g, h, pc) do { \ + uint32_t t1, t2; \ + SHA2_MEXP ## n(pc); \ + t1 = h + BSG2_1(e) + CH(e, f, g) \ + + K[pcount + (pc)] + W[(pc) & 0x0F]; \ + t2 = BSG2_0(a) + MAJ(a, b, c); \ + d += t1; \ + h = t1 + t2; \ + } while (0) + +#define SHA2_STEP1(a, b, c, d, e, f, g, h, pc) \ + SHA2_STEPn(1, a, b, c, d, e, f, g, h, pc) +#define SHA2_STEP2(a, b, c, d, e, f, g, h, pc) \ + SHA2_STEPn(2, a, b, c, d, e, f, g, h, pc) + + uint32_t A, B, C, D, E, F, G, H; + uint32_t W[16]; + unsigned pcount; + + A = val[0]; + B = val[1]; + C = val[2]; + D = val[3]; + E = val[4]; + F = val[5]; + G = val[6]; + H = val[7]; + pcount = 0; + SHA2_STEP1(A, B, C, D, E, F, G, H, 0); + SHA2_STEP1(H, A, B, C, D, E, F, G, 1); + SHA2_STEP1(G, H, A, B, C, D, E, F, 2); + SHA2_STEP1(F, G, H, A, B, C, D, E, 3); + SHA2_STEP1(E, F, G, H, A, B, C, D, 4); + SHA2_STEP1(D, E, F, G, H, A, B, C, 5); + SHA2_STEP1(C, D, E, F, G, H, A, B, 6); + SHA2_STEP1(B, C, D, E, F, G, H, A, 7); + SHA2_STEP1(A, B, C, D, E, F, G, H, 8); + SHA2_STEP1(H, A, B, C, D, E, F, G, 9); + SHA2_STEP1(G, H, A, B, C, D, E, F, 10); + SHA2_STEP1(F, G, H, A, B, C, D, E, 11); + SHA2_STEP1(E, F, G, H, A, B, C, D, 12); + SHA2_STEP1(D, E, F, G, H, A, B, C, 13); + SHA2_STEP1(C, D, E, F, G, H, A, B, 14); + SHA2_STEP1(B, C, D, E, F, G, H, A, 15); + for (pcount = 16; pcount < 64; pcount += 16) { + SHA2_STEP2(A, B, C, D, E, F, G, H, 0); + SHA2_STEP2(H, A, B, C, D, E, F, G, 1); + SHA2_STEP2(G, H, A, B, C, D, E, F, 2); + SHA2_STEP2(F, G, H, A, B, C, D, E, 3); + SHA2_STEP2(E, F, G, H, A, B, C, D, 4); + SHA2_STEP2(D, E, F, G, H, A, B, C, 5); + SHA2_STEP2(C, D, E, F, G, H, A, B, 6); + SHA2_STEP2(B, C, D, E, F, G, H, A, 7); + SHA2_STEP2(A, B, C, D, E, F, G, H, 8); + SHA2_STEP2(H, A, B, C, D, E, F, G, 9); + SHA2_STEP2(G, H, A, B, C, D, E, F, 10); + SHA2_STEP2(F, G, H, A, B, C, D, E, 11); + SHA2_STEP2(E, F, G, H, A, B, C, D, 12); + SHA2_STEP2(D, E, F, G, H, A, B, C, 13); + SHA2_STEP2(C, D, E, F, G, H, A, B, 14); + SHA2_STEP2(B, C, D, E, F, G, H, A, 15); + } + val[0] += A; + val[1] += B; + val[2] += C; + val[3] += D; + val[4] += E; + val[5] += F; + val[6] += G; + val[7] += H; +#endif +} + +static void +sha2small_update(br_sha224_context *cc, const void *data, size_t len) +{ + const unsigned char *buf; + size_t ptr; + + buf = data; + ptr = (size_t)cc->count & 63; + cc->count += (uint64_t)len; + while (len > 0) { + size_t clen; + + clen = 64 - ptr; + if (clen > len) { + clen = len; + } + memcpy(cc->buf + ptr, buf, clen); + ptr += clen; + buf += clen; + len -= clen; + if (ptr == 64) { + br_sha2small_round(cc->buf, cc->val); + ptr = 0; + } + } +} + +static void +sha2small_out(const br_sha224_context *cc, void *dst, int num) +{ + unsigned char buf[64]; + uint32_t val[8]; + size_t ptr; + + ptr = (size_t)cc->count & 63; + memcpy(buf, cc->buf, ptr); + memcpy(val, cc->val, sizeof val); + buf[ptr ++] = 0x80; + if (ptr > 56) { + memset(buf + ptr, 0, 64 - ptr); + br_sha2small_round(buf, val); + memset(buf, 0, 56); + } else { + memset(buf + ptr, 0, 56 - ptr); + } + br_enc64be(buf + 56, cc->count << 3); + br_sha2small_round(buf, val); + br_range_enc32be(dst, val, num); +} + +/* see bearssl.h */ +void +br_sha224_init(br_sha224_context *cc) +{ + cc->vtable = &br_sha224_vtable; + memcpy(cc->val, br_sha224_IV, sizeof cc->val); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_sha224_update(br_sha224_context *cc, const void *data, size_t len) +{ + sha2small_update(cc, data, len); +} + +/* see bearssl.h */ +void +br_sha224_out(const br_sha224_context *cc, void *dst) +{ + sha2small_out(cc, dst, 7); +} + +/* see bearssl.h */ +uint64_t +br_sha224_state(const br_sha224_context *cc, void *dst) +{ + br_range_enc32be(dst, cc->val, 8); + return cc->count; +} + +/* see bearssl.h */ +void +br_sha224_set_state(br_sha224_context *cc, const void *stb, uint64_t count) +{ + br_range_dec32be(cc->val, 8, stb); + cc->count = count; +} + +/* see bearssl.h */ +void +br_sha256_init(br_sha256_context *cc) +{ + cc->vtable = &br_sha256_vtable; + memcpy(cc->val, br_sha256_IV, sizeof cc->val); + cc->count = 0; +} + +/* see bearssl.h */ +void +br_sha256_out(const br_sha256_context *cc, void *dst) +{ + sha2small_out(cc, dst, 8); +} + +/* see bearssl.h */ +const br_hash_class br_sha224_vtable = { + sizeof(br_sha224_context), + BR_HASHDESC_ID(br_sha224_ID) + | BR_HASHDESC_OUT(28) + | BR_HASHDESC_STATE(32) + | BR_HASHDESC_LBLEN(6) + | BR_HASHDESC_MD_PADDING + | BR_HASHDESC_MD_PADDING_BE, + (void (*)(const br_hash_class **))&br_sha224_init, + (void (*)(const br_hash_class **, + const void *, size_t))&br_sha224_update, + (void (*)(const br_hash_class *const *, void *))&br_sha224_out, + (uint64_t (*)(const br_hash_class *const *, void *))&br_sha224_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_sha224_set_state +}; + +/* see bearssl.h */ +const br_hash_class br_sha256_vtable = { + sizeof(br_sha256_context), + BR_HASHDESC_ID(br_sha256_ID) + | BR_HASHDESC_OUT(32) + | BR_HASHDESC_STATE(32) + | BR_HASHDESC_LBLEN(6) + | BR_HASHDESC_MD_PADDING + | BR_HASHDESC_MD_PADDING_BE, + (void (*)(const br_hash_class **))&br_sha256_init, + (void (*)(const br_hash_class **, + const void *, size_t))&br_sha256_update, + (void (*)(const br_hash_class *const *, void *))&br_sha256_out, + (uint64_t (*)(const br_hash_class *const *, void *))&br_sha256_state, + (void (*)(const br_hash_class **, const void *, uint64_t)) + &br_sha256_set_state +}; |