// The input encrypted as though 128bit counter mode is being used. The extra
// state information to record how much of the 128bit block we have used is
// contained in *num, and the encrypted counter is kept in ecount_buf. Both
// *num and ecount_buf must be initialised with zeros before the first call to
// CRYPTO_ctr128_encrypt().
//
// This algorithm assumes that the counter is in the x lower bits of the IV
// (ivec), and that the application has full control over overflow and the rest
// of the IV. This implementation takes NO responsibility for checking that
// the counter doesn't overflow into the rest of the IV when incremented.
void CRYPTO_ctr128_encrypt(const uint8_t *in, uint8_t *out, size_t len,
const void *key, uint8_t ivec[16],
uint8_t ecount_buf[16], unsigned int *num,
block128_f block) {
unsigned int n;
assert(key && ecount_buf && num);
assert(len == 0 || (in && out));
assert(*num < 16);
n = *num;
while (n && len) {
*(out++) = *(in++) ^ ecount_buf[n];
--len;
n = (n + 1) % 16;
}
#if STRICT_ALIGNMENT
if (((uintptr_t)in | (uintptr_t)out |
(uintptr_t)ecount_buf) % sizeof(size_t) != 0) {
size_t l = 0;
while (l < len) {
if (n == 0) {
(*block)(ivec, ecount_buf, key);
ctr128_inc(ivec);
}
out[l] = in[l] ^ ecount_buf[n];
++l;
n = (n + 1) % 16;
}
*num = n;
return;
}
#endif
while (len >= 16) {
(*block)(ivec, ecount_buf, key);
ctr128_inc(ivec);
for (n = 0; n < 16; n += sizeof(size_t)) {
store_word_le(out + n,
load_word_le(in + n) ^ load_word_le(ecount_buf + n));
}
len -= 16;
out += 16;
in += 16;
n = 0;
}
if (len) {
(*block)(ivec, ecount_buf, key);
ctr128_inc(ivec);
while (len--) {
out[n] = in[n] ^ ecount_buf[n];
++n;
}
}
*num = n;
}
void CRYPTO_cbc128_encrypt(const uint8_t *in, uint8_t *out, size_t len,
const AES_KEY *key, uint8_t ivec[16],
block128_f block) {
size_t n;
const uint8_t *iv = ivec;
assert(key != NULL && ivec != NULL);
assert(len == 0 || (in != NULL && out != NULL));
if (STRICT_ALIGNMENT &&
((uintptr_t)in | (uintptr_t)out | (uintptr_t)ivec) % sizeof(size_t) !=
0) {
while (len >= 16) {
for (n = 0; n < 16; ++n) {
out[n] = in[n] ^ iv[n];
}
(*block)(out, out, key);
iv = out;
len -= 16;
in += 16;
out += 16;
}
} else {
while (len >= 16) {
for (n = 0; n < 16; n += sizeof(size_t)) {
store_word_le(out + n, load_word_le(in + n) ^ load_word_le(iv + n));
}
(*block)(out, out, key);
iv = out;
len -= 16;
in += 16;
out += 16;
}
}
while (len) {
for (n = 0; n < 16 && n < len; ++n) {
out[n] = in[n] ^ iv[n];
}
for (; n < 16; ++n) {
out[n] = iv[n];
}
(*block)(out, out, key);
iv = out;
if (len <= 16) {
break;
}
len -= 16;
in += 16;
out += 16;
}
OPENSSL_memcpy(ivec, iv, 16);
}
void CRYPTO_cbc128_decrypt(const uint8_t *in, uint8_t *out, size_t len,
const AES_KEY *key, uint8_t ivec[16],
block128_f block) {
size_t n;
union {
size_t t[16 / sizeof(size_t)];
uint8_t c[16];
} tmp;
assert(key != NULL && ivec != NULL);
assert(len == 0 || (in != NULL && out != NULL));
const uintptr_t inptr = (uintptr_t) in;
const uintptr_t outptr = (uintptr_t) out;
// If |in| and |out| alias, |in| must be ahead.
assert(inptr >= outptr || inptr + len <= outptr);
if ((inptr >= 32 && outptr <= inptr - 32) || inptr < outptr) {
// If |out| is at least two blocks behind |in| or completely disjoint, there
// is no need to decrypt to a temporary block.
const uint8_t *iv = ivec;
if (STRICT_ALIGNMENT &&
((uintptr_t)in | (uintptr_t)out | (uintptr_t)ivec) % sizeof(size_t) !=
0) {
while (len >= 16) {
(*block)(in, out, key);
for (n = 0; n < 16; ++n) {
out[n] ^= iv[n];
}
iv = in;
len -= 16;
in += 16;
out += 16;
}
} else if (16 % sizeof(size_t) == 0) { // always true
while (len >= 16) {
(*block)(in, out, key);
for (n = 0; n < 16; n += sizeof(size_t)) {
store_word_le(out + n, load_word_le(out + n) ^ load_word_le(iv + n));
}
iv = in;
len -= 16;
in += 16;
out += 16;
}
}
OPENSSL_memcpy(ivec, iv, 16);
} else {
// |out| is less than two blocks behind |in|. Decrypting an input block
// directly to |out| would overwrite a ciphertext block before it is used as
// the next block's IV. Decrypt to a temporary block instead.
if (STRICT_ALIGNMENT &&
((uintptr_t)in | (uintptr_t)out | (uintptr_t)ivec) % sizeof(size_t) !=
0) {
uint8_t c;
while (len >= 16) {
(*block)(in, tmp.c, key);
for (n = 0; n < 16; ++n) {
c = in[n];
out[n] = tmp.c[n] ^ ivec[n];
ivec[n] = c;
}
len -= 16;
in += 16;
out += 16;
}
} else if (16 % sizeof(size_t) == 0) { // always true
while (len >= 16) {
(*block)(in, tmp.c, key);
for (n = 0; n < 16; n += sizeof(size_t)) {
size_t c = load_word_le(in + n);
store_word_le(out + n,
tmp.t[n / sizeof(size_t)] ^ load_word_le(ivec + n));
store_word_le(ivec + n, c);
}
len -= 16;
in += 16;
out += 16;
}
}
}
while (len) {
uint8_t c;
(*block)(in, tmp.c, key);
for (n = 0; n < 16 && n < len; ++n) {
c = in[n];
out[n] = tmp.c[n] ^ ivec[n];
ivec[n] = c;
}
if (len <= 16) {
for (; n < 16; ++n) {
ivec[n] = in[n];
}
break;
}
len -= 16;
in += 16;
out += 16;
}
}
