Work in progress for Honda ECUs.

Usable but not completely verified. Some instruction decoding could be incomplete/invalid (I'm working on fixing them). Best way to use this project is to compare the output with asm662's. See Notes below for more information.

I used nios2.py and ht68fb560.py as basis for this project. It's my first time creating a processor module so there might be obvious mistake that I didn't see. Feel free to submit corrections. The target for this module is to be able to reverse 90's/00's Honda ECUs, more specifically Integra Type R's. So most testing will be done with such firmwares.

• honda-ecu-obd1-loader.py detects that a binary is a Honda ECU (should detect, at some point, when I know the header. Right now it accepts anything). This file should be added to the /loaders folder.
• oki-66207-processor.py is the processor file. oki66207.py is a temporary (?) split of the processor file to avoid having a big array in the main file, that slows down editing. Those files should be added to the /procs folder.
• The instruction array itself is built using ./helpers/op_to_array.py asm662-66207.op.modified. It started well, but it's a hack too in the end.

See TODO

If you know how to handle special processor flags in IDA/idapython, I'm curious (https://reverseengineering.stackexchange.com/q/22423/11827)

• This architecture uses a special register bit (DD flag) that affects instruction decoding. One byte could lead to two different instructions (similar, as it's basically "are we handling WORDs or BYTEs", but still different) depending on wether that flag is set or not. That flag can be set and reset at runtime which makes static decompilation pretty challenging. IDA performs multithreaded analysis so a global flag is not the solution. The current implementation keeps track of the DD flag for each instruction. Some testing revealed that the flag for some instructions is set/reset by different threads (xrefs are partially at fault), which shows that it's not a foolproof implementation, but it still does a ok-job.
• The concept of page addressing is hard to handle with only static analysis. LRB is the register that holds which page is referenced and its value can be changed anytime at runtime. For the same kind of reason DD flag is partly handled, "offset instructions" are decoded with the "off" keyword. As the exact value can't be computed 100% correctly, I leave this to the reverse engineer. When encountering the "off" keyword, please keep in mind that the value between the parenthesis is only the offset in the page.

• http://wuffs.org/blog/mouse-adventures-part-7
• https://github.com/markjandrews/idatraining/tree/master/exercises/exercise11_processor_module