Support for ST231

[GerdGattuso]

Is your feature request related to a problem? Please describe. I try to reverse a firmware for the ST231 processor (ST200 family). But didn't find a disassembler yet. Can you add support for ST231 to ghidra?

Describe the solution you'd like It would be nice if you can add the ST231 processor

Describe alternatives you've considered unfortunaly none

Additional context I attached the instruction set for the ST231. Here also two links:

http://www.audentia-gestion.fr/STMicroelectronics/PDF/en.CD17645929.pdf https://www.st.com/resource/en/technical_note/tn0614-high-performance-vliw-processor-core-stmicroelectronics.pdf

[mumbel]

build tools or sample binaries? This looks like a pretty complex ISA due to the bundles. Instruction decoding is trivia, they're all fairly simple instructions to do in SLEIGH, but implementing the SLEIGH where flow control is involved wont be straight forward

[GerdGattuso]

Build tools I unfortunately do not have. But I try to get the ST200 toolset. I have attached a sample binary. audio-firmware-38.3.0.24-4.zip

[mumbel]

Every time I look at this ISA I think I hate it more than the last time.

All of the instruction patterns need to track the stop bit for the bundles. With up to 4 instructions in a bundle, with the last instruction having the the bit set. op31=0 -> delayslot(1) op31=1 -> delayslot(0)

It seems the options then are:

1. if/goto

   if (op31:1 == 1) goto <no_delay>;
   delayslot(1);
   <no_delay>

2. attach bit pattern

   attach value [ op31 ] [ 1 0 ];
   ...
   delayslot(op31);

3. table

   stop: is op31=1 {}
   stop: is op31=0 { delayslot(1); }

It doesn't look like any approach really works that well:

• 3 I seem to get disassembly issues
• 2 just doesn't work, PCODE shows as UNIMPLEMENTED
• 1 seemed broken when I tried it (need to revisit).

extension immediate values are just awful. An instruction can use a signed 9-bit immediate, but if after immr or before imml in a bundle, that becomes an unsigned and added to the 23-bit extension (shifted 9 bits). Getting the extension seems to be just a case of where the delayslot(1) is called (before or after reading the extension operand); but changing the signed to unsigned 9-bit immediate does not appear to be straight forward.

[mumbel]

https://github.com/mumbel/ST200

I don't intend on any further development, but figured I'd post my attempt. Bundles don't seem to work with the current delayslot() from the things I've tried.

[GerdGattuso]

Okay. Thank you for your commitment. I try to get the ST200 Toolset. Maybe the Toolset (with compiler) will be helpful.