/*
 * base16.h -- Fast Base16 stream decoder
 *
 * Copyright (c) 2005-2016, Nick Galbreath.
 * Copyright (c) 2022, NLnet Labs. All rights reserved.
 *
 * SPDX-License-Identifier: BSD-2-Clause
 *
 */
#ifndef BASE16_H
#define BASE16_H

// adaptation of base16 decoder by Nick Galbreath to operate similar to the
// base64 stream decoder by Alfred Klomp.
// https://github.com/client9/stringencoders
// https://github.com/aklomp/base64

struct base16_state {
  int eof;
  int bytes;
  unsigned char carry;
};

#define BASE16_EOF 1

static const uint32_t base16_table_dec_32bit_d0[256] = {
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 257, 256,
  0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 256, 256,
  256, 256, 256, 256, 256, 160, 176, 192, 208, 224, 240, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 160, 176, 192, 208, 224, 240, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256
};

static const uint32_t base16_table_dec_32bit_d1[256] = {
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 257, 256,
  0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 256, 256,
  256, 256, 256, 256, 256, 10, 11, 12, 13, 14, 15, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 10, 11, 12, 13, 14, 15, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256, 256,
  256, 256, 256, 256
};

static really_inline int
base16_dec_loop_generic_32_inner(
  const uint8_t **s, uint8_t **o, size_t *rounds)
{
  const uint32_t val1 = base16_table_dec_32bit_d0[(*s)[0]]
                      | base16_table_dec_32bit_d1[(*s)[1]];
  const uint32_t val2 = base16_table_dec_32bit_d0[(*s)[2]]
                      | base16_table_dec_32bit_d1[(*s)[3]];

  if (val1 > 0xff || val2 > 0xff)
    return 0;

  (*o)[0] = (uint8_t)val1;
  (*o)[1] = (uint8_t)val2;

  *s += 4;
  *o += 2;
  *rounds -= 1;

  return 1;
}

static really_inline void
base16_dec_loop_generic_32(
  const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
{
  if (*slen < 4)
    return;

  // some comment on the how and what...
  size_t rounds = (*slen - 4) / 4;

  *slen -= rounds * 4; // 4 bytes consumed per round
  *olen += rounds * 2; // 2 bytes produced per round

  do {
    if (rounds >= 8) {
      if (base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds))
        continue;
      break;
    }
    if (rounds >= 4) {
      if (base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds))
        continue;
      break;
    }
    if (rounds >= 2) {
      if (base16_dec_loop_generic_32_inner(s, o, &rounds) &&
          base16_dec_loop_generic_32_inner(s, o, &rounds))
        continue;
      break;
    }
    base16_dec_loop_generic_32_inner(s, o, &rounds);
    break;
  } while (rounds > 0);

  // Adjust for any rounds that were skipped:
  *slen += rounds * 4;
  *olen -= rounds * 2;
}

nonnull((1,2,4,5))
static really_inline int base16_stream_decode(
  struct base16_state *state,
  const char *src,
  size_t srclen,
  uint8_t *out,
  size_t *outlen)
{
  int ret = 0;
  const uint8_t *s = (const uint8_t *) src;
  uint8_t *o = (uint8_t *) out;
  uint32_t q;

  // Use local temporaries to avoid cache thrashing:
  size_t olen = 0;
  size_t slen = srclen; 
  struct base16_state st;
  st.eof = state->eof;
  st.bytes = state->bytes;
  st.carry = state->carry;

  if (st.eof) {
    *outlen = 0;
    return ret;
  }

  // Duff's device again:
  switch (st.bytes)
  {
#if defined(__SUNPRO_C)
#pragma error_messages(off, E_STATEMENT_NOT_REACHED)
#endif
    for (;;)
#if defined(__SUNPRO_C)
#pragma error_messages(default, E_STATEMENT_NOT_REACHED)
#endif
    {
    case 0:
      base16_dec_loop_generic_32(&s, &slen, &o, &olen);
      if (slen-- == 0) {
        ret = 1;
        break;
      }
      if ((q = base16_table_dec_32bit_d0[*s++]) >= 255) {
        st.eof = BASE16_EOF;
        break;
      }
      st.carry = (uint8_t)q;
      st.bytes = 1;

      // fallthrough

    case 1:
      if (slen-- == 0) {
        ret = 1;
        break;
      }
      if ((q = base16_table_dec_32bit_d1[*s++]) >= 255) {
        st.eof = BASE16_EOF;
        break;
      }
      *o++ = st.carry | (uint8_t)q;
      st.carry = 0;
      st.bytes = 0;
      olen++;
    }
  }

  state->eof = st.eof;
  state->bytes = st.bytes;
  state->carry = st.carry;
  *outlen = olen;
  return ret;
}

nonnull((1,3,4))
static really_inline int base16_decode(
  const char *src, size_t srclen, uint8_t *out, size_t *outlen)
{
  struct base16_state state = { .eof = 0, .bytes = 0, .carry = 0 };
  return base16_stream_decode(&state, src, srclen, out, outlen) & !state.bytes;
}

// FIXME: RFC3597 section 5 states each word of data must contain an even
//        number of hexadecimal digits. The same is not true for DS records.
//        RFC4043 section 5.3 merely states whitespace is allowed within the
//        hexadecimal text. Words containing an uneven number of hexadecimal
//        digits are impractical, but supported (BIND supports uneven
//        sequences)
nonnull_all
static really_inline int32_t parse_base16_sequence(
  parser_t *parser,
  const type_info_t *type,
  const rdata_info_t *field,
  rdata_t *rdata,
  token_t *token)
{
  if (is_contiguous(token)) {
    struct base16_state state = { .eof = 0, .bytes = 0, .carry = 0 };

    do {
      size_t length = (token->length + 1) / 2;
      if ((uintptr_t)rdata->limit - (uintptr_t)rdata->octets < length)
        SYNTAX_ERROR(parser, "Invalid %s in %s", NAME(field), NAME(type));
      if (!base16_stream_decode(&state, token->data, token->length, rdata->octets, &length))
        SYNTAX_ERROR(parser, "Invalid %s in %s", NAME(field), NAME(type));
      rdata->octets += length;
      take(parser, token);
    } while (is_contiguous(token));

    if (state.bytes)
      *rdata->octets++ = state.carry;
  }

  return have_delimiter(parser, type, token);
}

nonnull_all
static really_inline int32_t parse_base16(
  parser_t *parser,
  const type_info_t *type,
  const rdata_info_t *field,
  rdata_t *rdata,
  const token_t *token)
{
  size_t length = token->length / 2;
  if ((uintptr_t)rdata->limit - (uintptr_t)rdata->octets < length)
    SYNTAX_ERROR(parser, "Invalid %s in %s", NAME(field), NAME(type));
  if (!base16_decode(token->data, token->length, rdata->octets, &length))
    SYNTAX_ERROR(parser, "Invalid %s in %s", NAME(field), NAME(type));
  rdata->octets += length;
  return 0;
}

nonnull_all
static really_inline int32_t parse_salt(
  parser_t *parser,
  const type_info_t *type,
  const rdata_info_t *field,
  rdata_t *rdata,
  const token_t *token)
{
  if (token->length == 1 && token->data[0] == '-')
    return (void)(*rdata->octets++ = 0), 0;

  size_t length = token->length / 2;
  uint8_t *octets = rdata->octets++;
  // FIXME: not quite right yet! we must not exceed 255 octets!
  if ((uintptr_t)rdata->limit - (uintptr_t)rdata->octets < (length + 1))
    SYNTAX_ERROR(parser, "Invalid %s in %s", NAME(field), NAME(type));
  if (!base16_decode(token->data, token->length, rdata->octets, &length))
    SYNTAX_ERROR(parser, "Invalid %s in %s", NAME(field), NAME(type));
  *octets = (uint8_t)length;
  rdata->octets += length;
  return 0;
}

#endif // BASE16_H