firmware/.clusterfuzzlite/README.md
Eric Severance b0fe5ef8ba
Initial commit of a fuzzer for Meshtastic (#5790)
* Initial commit of a fuzzer for Meshtastic.

* Use a max of 5 for the phone queues

* Only write files to the temp dir

* Limitless queue + fuzzer = lots of ram :)

* Use $PIO_ENV for path to program

* spelling: s/is/to/

* Use loopCanSleep instead of a lock in Router

* realHardware allows full use of a CPU core

* Ignore checkov CKV_DOCKER_2 & CKV_DOCKER_3

* Add Atak seed

* Fix lint issues in build.sh

* Use exception to exit from portduino_main

* Separate build & source files into $WORK & $SRC

* Use an ephemeral port for the API server

* Include CXXFLAGS in the link step

* Read all shared libraries

* Use a separate work directory for each sanitizer

---------

Co-authored-by: Ben Meadors <benmmeadors@gmail.com>
2025-01-16 18:42:21 -06:00

60 lines
2.5 KiB
Markdown

# ClusterFuzzLite for Meshtastic
This directory contains the fuzzer implementation for Meshtastic using the ClusterFuzzLite framework.
See the [ClusterFuzzLite documentation](https://google.github.io/clusterfuzzlite/) for more details.
## Running locally
ClusterFuzzLite uses the OSS-Fuzz toolchain. To build the fuzzer manually, first grab a copy of OSS-Fuzz.
```shell
git clone https://github.com/google/oss-fuzz.git
cd oss-fuzz
```
To build the fuzzer, run:
```shell
python3 infra/helper.py build_image --external $PATH_TO_MESHTASTIC_FIRMWARE_DIRECTORY
python3 infra/helper.py build_fuzzers --external $PATH_TO_MESHTASTIC_FIRMWARE_DIRECTORY --sanitizer address
```
To run the fuzzer, run:
```shell
python3 infra/helper.py run_fuzzer --external --corpus-dir=<path-to-temp-corpus-dir> $PATH_TO_MESHTASTIC_FIRMWARE_DIRECTORY router_fuzzer
```
More background on these commands can be found in the
[ClusterFuzzLite documentation](https://google.github.io/clusterfuzzlite/build-integration/#testing-locally).
## router_fuzzer.cpp
This fuzzer submits MeshPacket protos to the `Router::enqueueReceivedMessage` method. It takes the binary
data from the fuzzer and decodes that data to a MeshPacket using nanopb. A few fields in
the MeshPacket are modified by the fuzzer.
- If the `to` field is 0, it will be replaced with the NodeID of the running node.
- If the `from` field is 0, it will be replaced with the NodeID of the running node.
- If the `id` field is 0, it will be replaced with an incrementing counter value.
- If the `pki_encrypted` field is true, the `public_key` field will be populated with the first admin key.
The `router_fuzzer_seed_corpus.py` file contains a list of MeshPackets. It is run from inside build.sh and
writes the binary MeshPacket protos to files. These files are use used by the fuzzer as its initial seed data,
helping the fuzzer to start off with a few known inputs.
### Interpreting a fuzzer crash
If the fuzzer crashes, it'll write the input bytes used for the test case to a file and notify about the
location of that file. The contents of the file are a binary serialized MeshPacket protobuf. The following
snippet of Python code can be used to parse the file into a human readable form.
```python
from meshtastic.protobuf import mesh_pb2
mesh_pb2.MeshPacket.FromString(open("crash-XXXX-file", "rb").read())
```
Consider adding any such crash results to the `router_fuzzer_seed_corpus.py` file to ensure there a isn't
a future regression for that crash test case.