Files
cryptsetup/tests/fuzz
Milan Broz 661f57def4 Use SPDX license identifiers.
This patch switches code to SPDX one-line license identifiers according to
https://spdx.dev/learn/handling-license-info/
and replacing long license text headers.

I used C++ format on the first line in style
// SPDX-License-Identifier: <id>
except exported libcryptsetup.h, when only C comments are used.

The only additional changes are:
- switch backend utf8.c from LGPL2+ to LGPL2.1+ (as in systemd)
- add some additional formatting lines.
2024-06-03 16:38:15 +00:00
..
2024-06-03 16:38:15 +00:00
2023-02-01 16:02:26 +01:00
2022-10-05 09:49:55 +02:00

Fuzzing target for cryptsetup project

This directory contains experimental targets for fuzzing testing. It can be run in the OSS-Fuzz project but also compiled separately.

Requirements

Fuzzers use address sanitizer. To properly detect problems, all important libraries must be compiled statically with sanitizer enabled.

Compilation requires clang and clang++ compilers (gcc is not supported yet).

Standalone build

The script oss-fuzz-build.sh can be used to prepare the tree with pre-compiled library dependencies. We use upstream git for projects, which can clash with locally installed versions. The best is to use only basic system installation without development packages (script will use custom include, libs, and pkg-config paths).

Build Docker image and fuzzers

You can also run OSS-Fuzz in a Docker image, use these commands to prepare fuzzers:

sudo python3 infra/helper.py build_image cryptsetup
sudo python3 infra/helper.py build_fuzzers cryptsetup

On SELinux systems also add (https://github.com/google/oss-fuzz/issues/30):

sudo chcon -Rt svirt_sandbox_file_t build/

Run LUKS2 fuzzer

FUZZER_NAME can be one of: crypt2_load_fuzz, crypt2_load_proto_fuzz, crypt2_load_proto_plain_json_fuzz

FUZZER_NAME="crypt2_load_proto_plain_json_fuzz"
sudo mkdir -p build/corpus/cryptsetup/$FUZZER_NAME
sudo python infra/helper.py run_fuzzer --corpus-dir build/corpus/cryptsetup/$FUZZER_NAME/ --sanitizer address cryptsetup $FUZZER_NAME '-jobs=8 -workers=8'

The output of the parallel threads will be written to fuzz-<N>.log (where <N> is the number of the process). You can watch it using e.g.:

tail -f build/out/cryptsetup/fuzz-*

Optionally, you can use experimental fork mode for parallelization and the output will be displayed directly on the terminal:

sudo python infra/helper.py run_fuzzer --corpus-dir build/corpus/cryptsetup/$FUZZER_NAME/ --sanitizer address cryptsetup $FUZZER_NAME '-fork=8 '

Rebuild fuzz targets for coverage

sudo python infra/helper.py build_fuzzers --sanitizer coverage cryptsetup

Generate coverage report

sudo python infra/helper.py coverage cryptsetup --no-corpus-download --fuzz-target $FUZZER_NAME

Further information

For more details, you can look into the Using fuzzing for Linux disk encryption tools thesis.