CHACHA_BLOCKLEN
u8 ks[CHACHA_BLOCKLEN];
KKASSERT(in_len % CHACHA_BLOCKLEN == 0); /* must be complete blocks */
KKASSERT(in_len <= CHACHA_BLOCKLEN);
len = rounddown2(src_len, CHACHA_BLOCKLEN);
len = rounddown2(data_len, CHACHA_BLOCKLEN);
uint8_t block[CHACHA_BLOCKLEN], *p;