Skip to content

Latest commit

 

History

History
332 lines (243 loc) · 11.7 KB

HACKING.md

File metadata and controls

332 lines (243 loc) · 11.7 KB
Raw

Hacking

Welcome, fellow coder!

This repository contains a code to run a lightning protocol daemon. It's broken into subdaemons, with the idea being that we can add more layers of separation between different clients and extra barriers to exploits.

It is designed to implement the lightning protocol as specified in various BOLTs.

Getting Started

It's in C, to encourage alternate implementations. Patches are welcome! You should read our Style Guide.

To read the code, you should start from lightningd.c and hop your way through the '~' comments at the head of each daemon in the suggested order.

The Components

Here's a list of parts, with notes:

  • ccan - useful routines from http://ccodearchive.net

    • Use make update-ccan to update it.
    • Use make update-ccan CCAN_NEW="mod1 mod2..." to add modules
    • Do not edit this! If you want a wrapper, add one to common/utils.h.

  • bitcoin/ - bitcoin script, signature and transaction routines.

    • Not a complete set, but enough for our purposes.

  • external/ - external libraries from other sources

    • libbacktrace - library to provide backtraces when things go wrong.
    • libsodium - encryption library (should be replaced soon with built-in)
    • libwally-core - bitcoin helper library
    • secp256k1 - bitcoin curve encryption library within libwally-core
    • jsmn - tiny JSON parsing helper

  • tools/ - tools for building

    • check-bolt.c: check the source code contains correct BOLT quotes (as used by check-source)
    • generate-wire.py: generates wire marshal/unmarshal-ing routines for subdaemons and BOLT specs.
    • mockup.sh / update-mocks.sh: tools to generate mock functions for unit tests.

  • tests/ - blackbox tests (mainly)

    • unit tests are in tests/ subdirectories in each other directory.

  • doc/ - you are here

  • devtools/ - tools for developers

    • Generally for decoding our formats.

  • contrib/ - python support and other stuff which doesn't belong :)

  • wire/ - basic marshalling/un for messages defined in the BOLTs

  • common/ - routines needed by any two or more of the directories below

  • cli/ - commandline utility to control lightning daemon.

  • lightningd/ - master daemon which controls the subdaemons and passes peer file descriptors between them.

  • wallet/ - database code used by master for tracking what's happening.

  • hsmd/ - daemon which looks after the cryptographic secret, and performs commitment signing.

  • gossipd/ - daemon to maintain routing information and broadcast gossip.

  • connectd/ - daemon to connect to other peers, and receive incoming.

  • openingd/ - daemon to open a channel for a single peer, and chat to a peer which doesn't have any channels/

  • channeld/ - daemon to operate a single peer once channel is operating normally.

  • closingd/ - daemon to handle mutual closing negotiation with a single peer.

  • onchaind/ - daemon to handle a single channel which has had its funding transaction spent.

Debugging

You can build Core Lightning with DEVELOPER=1 to use dev commands listed in cli/lightning-cli help. ./configure --enable-developer will do that. You can log console messages with log_info() in lightningd and status_debug() in other subdaemons.

You can debug crashing subdaemons with the argument --dev-debugger=channeld, where channeld is the subdaemon name. It will run gnome-terminal by default with a gdb attached to the subdaemon when it starts. You can change the terminal used by setting the DEBUG_TERM environment variable, such as DEBUG_TERM="xterm -e" or DEBUG_TERM="konsole -e".

It will also print out (to stderr) the gdb command for manual connection. The subdaemon will be stopped (it sends itself a SIGSTOP); you'll need to continue in gdb.

Database

Core Lightning state is persisted in lightning-dir. It is a sqlite database stored in the lightningd.sqlite3 file, typically under ~/.lightning/<network>/. You can run queries against this file like so:

$ sqlite3 ~/.lightning/bitcoin/lightningd.sqlite3 \
  "SELECT HEX(prev_out_tx), prev_out_index, status FROM outputs"

Or you can launch into the sqlite3 repl and check things out from there:

$ sqlite3 ~/.lightning/bitcoin/lightningd.sqlite3
SQLite version 3.21.0 2017-10-24 18:55:49
Enter ".help" for usage hints.
sqlite> .tables
channel_configs  invoices         peers            vars
channel_htlcs    outputs          shachain_known   version
channels         payments         shachains
sqlite> .schema outputs
...

Some data is stored as raw bytes, use HEX(column) to pretty print these.

Make sure that clightning is not running when you query the database, as some queries may lock the database and cause crashes.

Common variables

Table vars contains global variables used by lightning node.

$ sqlite3 ~/.lightning/bitcoin/lightningd.sqlite3
SQLite version 3.21.0 2017-10-24 18:55:49
Enter ".help" for usage hints.
sqlite> .headers on
sqlite> select * from vars;
name|val
next_pay_index|2
bip32_max_index|4
...

Variables:

  • next_pay_index next resolved invoice counter that will get assigned.
  • bip32_max_index last wallet derivation counter.

Note: Each time newaddr command is called, bip32_max_index counter is increased to the last derivation index. Each address generated after bip32_max_index is not included as lightning funds.

Build and Development

Install the following dependencies for best results:

sudo apt update
sudo apt install valgrind cppcheck shellcheck libsecp256k1-dev libpq-dev

Re-run configure and build using make:

./configure --enable-developer
make -j$(nproc)

Testing

Tests are run with: make check [flags] where the pertinent flags are:

DEVELOPER=[0|1] - developer mode increases test coverage
VALGRIND=[0|1]  - detects memory leaks during test execution but adds a significant delay
PYTEST_PAR=n    - runs pytests in parallel

A modern desktop can build and run through all the tests in a couple of minutes with:

make -j12 full-check PYTEST_PAR=24 DEVELOPER=1 VALGRIND=0

Adjust -j and PYTEST_PAR accordingly for your hardware.

There are four kinds of tests:

  • source tests - run by make check-source, looks for whitespace, header order, and checks formatted quotes from BOLTs if BOLTDIR exists.

  • unit tests - standalone programs that can be run individually. You can also run all of the unit tests with make check-units. They are run-*.c files in test/ subdirectories used to test routines inside C source files.

    You should insert the lines when implementing a unit test:

    /* AUTOGENERATED MOCKS START */

    /* AUTOGENERATED MOCKS END */

    and make update-mocks will automatically generate stub functions which will allow you to link (and conveniently crash if they're called).

  • blackbox tests - These tests setup a mini-regtest environment and test lightningd as a whole. They can be run individually:

    PYTHONPATH=contrib/pylightning:contrib/pyln-client:contrib/pyln-testing:contrib/pyln-proto py.test -v tests/

    You can also append -k TESTNAME to run a single test. Environment variables DEBUG_SUBD=<subdaemon> and TIMEOUT=<seconds> can be useful for debugging subdaemons on individual tests.

  • pylightning tests - will check contrib pylightning for codestyle and run the tests in contrib/pylightning/tests afterwards:

    make check-python

Our Github Actions instance (see .github/workflows/*.yml) runs all these for each pull request.

Additional Environment Variables

TEST_CHECK_DBSTMTS=[0|1]            - When running blackbox tests, this will
                                      load a plugin that logs all compiled
                                      and expanded database statements.
                                      Note: Only SQLite3.
TEST_DB_PROVIDER=[sqlite3|postgres] - Selects the database to use when running
                                      blackbox tests.
EXPERIMENTAL_DUAL_FUND=[0|1]	    - Enable dual-funding tests.

Troubleshooting

Valgrind complains about code we don't control

Sometimes valgrind will complain about code we do not control ourselves, either because it's in a library we use or it's a false positive. There are generally three ways to address these issues (in descending order of preference):

  1. Add a suppression for the one specific call that is causing the issue. Upon finding an issue valgrind is instructed in the testing framework to print filters that'd match the issue. These can be added to the suppressions file under tests/valgrind-suppressions.txt in order to explicitly skip reporting these in future. This is preferred over the other solutions since it only disables reporting selectively for things that were manually checked. See the valgrind docs for details.
  2. Add the process that valgrind is complaining about to the --trace-children-skip argument in pyln-testing. This is used in cases of full binaries not being under our control, such as the python3 interpreter used in tests that run plugins. Do not use this for binaries that are compiled from our code, as it tends to mask real issues.
  3. Mark the test as skipped if running under valgrind. It's mostly used to skip tests that otherwise would take considerably too long to test on CI. We discourage this for suppressions, since it is a very blunt tool.

Making BOLT Modifications

All of code for marshalling/unmarshalling BOLT protocol messages is generated directly from the spec. These are pegged to the BOLTVERSION, as specified in Makefile.

Source code analysis

An updated version of the NCC source code analysis tool is available at

https://github.com/bitonic-cjp/ncc

It can be used to analyze the lightningd source code by running make clean && make ncc. The output (which is built in parallel with the binaries) is stored in .nccout files. You can browse it, for instance, with a command like nccnav lightningd/lightningd.nccout.

Subtleties

There are a few subtleties you should be aware of as you modify deeper parts of the code:

  • ccan/structeq's STRUCTEQ_DEF will define safe comparison function foo_eq() for struct foo, failing the build if the structure has implied padding.
  • command_success, command_fail, and command_fail_detailed will free the cmd you pass in. This also means that if you tal-allocated anything from the cmd, they will also get freed at those points and will no longer be accessible afterwards.
  • When making a structure part of a list, you will instance a struct list_node. This has to be the first field of the structure, or else dev-memleak command will think your structure has leaked.

Protocol Modifications

The source tree contains CSV files extracted from the v1.0 BOLT specifications (wire/extracted_peer_wire_csv and wire/extracted_onion_wire_csv). You can regenerate these by first deleting the local copy(if any) at directory .tmp.lightning-rfc, setting BOLTDIR and BOLTVERSION appropriately, and finally running make extract-bolt-csv. By default the bolts will be retrieved from the directory ../bolts and a recent git version.

e.g., make extract-bolt-csv BOLTDIR=../bolts BOLTVERSION=ee76043271f79f45b3392e629fd35e47f1268dc8

Further Information

Feel free to ask questions on the lightning-dev mailing list, or on #c-lightning on IRC, or email me at rusty@rustcorp.com.au.

Cheers!
Rusty.