libfreefare/HACKING.md

101 lines
4.5 KiB
Markdown
Raw Normal View History

Hello hackers!
# General remarks about contributing
Contributions to the libfreefare are welcome! Here are some directions to get
you started:
## Install Cutter
[Cutter](http://cutter.sourceforge.net/) is a cool unit test framework for C
code. You will need it to run the regression tests suite (`make check`) and
ensure that your code does not break other features.
To cover all tests you'll need several blank cards:
* MIFARE Classic 4k with default keys (`FFFFFFFFFF`)
* MIFARE DESFire EV1 4k with default PICC key (`0000000000000000`)
* MIFARE Ultralight
* MIFARE Ultralight C with default key (`BREAKMEIFYOUCAN!`)
After "make check", you can run sub-sets of tests directly with cutter:
~~~
$ cutter -n /ultralight/ test
$ cutter -n /classic/ test
$ cutter -n /desfire/ test
~~~
## Follow style conventions
The source code of the library trend to follow some conventions so that it is
consistent in style and thus easier to read. Basically, it follows [FreeBSD's
style(9)](http://www.freebsd.org/cgi/man.cgi?query=style); adding 4-space
indentation and 8-space tabs (which you should configure in your editor, e.g.
`:set sw=4 ts=8` in vim). You are also advised to `:set cinoptions=t0(0:0` so
that you don't have to care about indentation anymore. For more details, please
read the [style(9) man page from FreeBSD's
website](http://www.freebsd.org/cgi/man.cgi?query=style).
For style correction install package `astyle`. Then run `make style`.
## Write tests
I already told you cutter is lovely, so you really should use it! If you want
to contribute code, write test: only code with appropriate tests will be
considered. And remember that
[TDD](http://en.wikipedia.org/wiki/Test-driven_development) (Test Driven
Development) is cool and writing all tests at the end deeply depressing, so
test early, test often!
# Adding support for a new type of card
Adding a new supported card to the libfreefare requires a few modification in
multiple places. Here is a list of the things to do in order to have the
infrastructure ready for hacking the new card support:
- Edit `libfreefare/freefare.h`:
- Add your tag to the `freefare_tag_type` enum;
- Add a `<tag>_connect()` and a `<tag>_disconnect()` function prototype;
- Edit `libfreefare/freefare.3`:
- Add your tag to the `freefare_tag_type' enum documentation;
- Edit `libfreefare/freefare_internal.h`:
- Add a new `<tag>_tag struct`. It's very first member SHALL be `struct
freefare_tag __tag`;
- Add a `<tag>_tag_new()` and a `<tag>_tag_free()` function prototype;
- Add a `ASSERT_<TAG>()` macro to check the tag's type;
- Add a `<TAG>()` macro to cast a generic tag to your type.
- Edit `libfreefare/freefare.c`:
- Add your tag type to the supported_tags array;
- Edit the `freefare_get_tags()` function so that it calls `<tag>_tag_new()`
when it finds your tag;
- Edit the `freefare_free_tags()` function so that it calls
`<tag>_tag_free()` to free your tags;
- Create `libfreefare/<tag>.c` and implement all that's missing:
- `<tag>_tag_new()` MUST allocate all data-structure the tag may need to
use during it's lifetime. We do not want to have any function to fail
later because the running system is out of resources. Buffers for
cryptographic operations on random amount of data MAY however be
(re)allocated on demand, in such case refrain from shrinking
unnecessarily the buffer size.
- `<tag>_connect()` SHOULD initialise data allocated by `<tag>_tag_new()`.
Keep in mind that a single tag may be disconnected from and connected
to again, without being freed in the meantime. Since all memory
allocations are done in `<tag>_tag_new()`, your code SHOULD only care
about initialising these data structures;
- `<tag>_disconnect()` MAY do more that just send a disconnect command to
the tag. At time of writing I have no idea what it could be but who
knows...
- `<tag>_tag_free()` SHALL free all resources allocated for the tag
(surprising, isn't it?)
# Various guidelines
- If a given card has different cryptographic modes, you SHOULD use
switch/cases to handle specific branches of code, even when applicable to
only one cypher. The idea is that if you don't provide support for all
cryptographic schemes, or if an evolution of the card provides more
cryptographic possibilities, when adding support for a new cypher, the
compiler can warn the developer about unhandled values in switch
statements. Please refer to the MIFARE DESFire code for an example.