Making a basic TLV parser in Kaitai

Question

I can't figure out how to write a simple TLV definition in Kaitai Struct.

I reduced the problem to a simple TLV file where the following repeats until the end of file:

• One byte for the tag
• One byte for the lenght, in bytes
• That many bytes

This definition:

meta:
  id: basictlv
  endian: le
types:
  bunch_of_ones:
    seq:
      - id: how_many_1
        type: u1
      - id: ones
        size: how_many_1
  bunch_of_twos:
    seq:
      - id: how_many_2
        type: u1
      - id: twos
        size: how_many_2
seq:
  - id: type_tag
    type: u1
  - id: body
    type:
      switch-on: type_tag
      cases:
        1: bunch_of_ones
        2: bunch_of_twos

Used on this file

$ xxd simple-tlv-file.dat
00000000: 0102 0101 0203 0202 0203 0103            ............

Switches on the type, but it only parse the first TLV and does not repeat.

$ ksdump simple-tlv-file.dat sample-tlv.yaml 2> /dev/null
body:
  how_many_1: 2
  ones: 01 01
type_tag: 1

Answer 1

To parse a basic TLV with Kaitai Struct, you must break down all the possibilities as types, and in your main sequence (aka seq) you read the top level "object" until eos (or some other mean to mark its end). This way, you can also add a header or other data in the file before the actual TLV structure.

So in your example, move the TLV block to the types section:

types:
  block_of_data:
    seq:
    - id: type_tag
      type: u1
    - id: body
      type:
        switch-on: type_tag
        cases:
          1: bunch_of_ones
          2: bunch_of_twos

You can also add another type, like a header to the types section

  hello_header:
    seq:
      - id: magic
        contents:
          - 0x48
          - 0x45
          - 0x4C
          - 0x4C
          - 0x4F

And you will be able to parse this file:

$ xxd ~/downloads/simple-tlv-file.dat
00000000: 4845 4c4c 4f02 0101 0203 0202 0201 0201  HELLO...........
00000010: 01

With a main seq like this one:

seq:
  - id: header
    type: hello_header
  - id: data
    type: block_of_data
    repeat: eos