Re: [babel] Mandatory sub-TLVs in Next Hop and Router-ID

[Toke Høiland-Jørgensen <[e-mail]>]

Juliusz Chroboczek <[e-mail]> writes:

>> I suppose that if you're building a parse tree, then the two are roughly
>> similar.
>
> Hmm, I'm no longer so sure about that; it might actually be
> significantly simpler. Is that your case, Toke?

I'm not building a parse tree, I'm just doing two-pass parsing (first
pass reads wire data into an internal data structure while updating
parser state, second pass does protocol handling). The pseudo-code
example I posted before is not actually that far off from the actual C
function:

static inline int
babel_read_tlv(struct babel_tlv *hdr,
               union babel_msg *msg,
               struct babel_parse_state *state)
{
  if ((hdr->type <= BABEL_TLV_PADN) ||
      (hdr->type >= BABEL_TLV_MAX) ||
      !tlv_data[hdr->type].read_tlv)
    return PARSE_IGNORE;

  if (TLV_LENGTH(hdr) < tlv_data[hdr->type].min_length)
    return PARSE_ERROR;

  memset(msg, 0, sizeof(*msg));
  return tlv_data[hdr->type].read_tlv(hdr, msg, state);
}


What I was planning to do was just add a babel_read_subtlv() function
that would be called after the last line, which could just return
PARSE_IGNORE on unknown mandatory sub-TLVs (which would cause the
first-pass parser to drop the whole TLV).


I know this is very specific to the way I implemented things, but from
a high-level PoV, I'd agree with Markus that being consistent is better
than special-casing.

-Toke