4 releases
0.1.3 | Apr 11, 2022 |
---|---|
0.1.2 | Apr 11, 2022 |
0.1.1 | Apr 11, 2022 |
0.1.0 | Apr 11, 2022 |
#391 in Testing
470KB
2K
SLoC
Fazi
A reimplementation of libfuzzer in Rust with some improvements
Supported Features
- libFuzzer's mutations
- SanCov feedback
- Building without a
main()
entry point - Crash replaying
- Recoverage
- Custom user initialization
- Scaling/forking support
- Timeout detection
- Custom dictionaries
- Custom user-defined mutations
Anything else that's missing from libFuzzer's featureset will likely not be supported. Feel free to file an issue if you'd like to voice support for something.
Usage
Step 1: Build fazi (requires Rust nightly):
$ cargo build --release
Note: Fazi can also be built without the main entrypoint by providing the --no-default-features
flag
Step 2: Build your harness:
$ FAZI_DIR="../path-to-fazi" clang ./main.c -fsanitize=fuzzer-no-link -fsanitize=address -lfazi -L$FAZI_DIR/target/release/
Step 3: Run the harness:
$ ./a.out
You can list command-line options with the --help
flag:
fazi
USAGE:
a.out [OPTIONS] [SUBCOMMAND]
OPTIONS:
--corpus-dir <CORPUS_DIR>
Location at which inputs that cause new coverage will be saved [default: ./corpus]
--crashes-dir <CRASHES_DIR>
Location at which crashing inputs will be saved [default: ./crashes]
-h, --help
Print help information
--len-control <LEN_CONTROL>
Length control is used in an algorithm for deciding how quickly the input size grows. A
larger value will result in faster growth while a smaller value will result in slow
growth [default: 100]
--max-input-len <MAX_INPUT_LEN>
The maximum size (in bytes) that an input can extend to [default: 65000]
--max-iters <MAX_ITERS>
Maximum number of fuzzing iterations before the fuzzer should exit
--max-mutation-depth <MAX_MUTATION_DEPTH>
The maximum number of times to mutate a single input before moving on to another
[default: 15]
--seed <SEED>
RNG seed
SUBCOMMANDS:
help Print this message or the help of the given subcommand(s)
repro Reproduce some crash
Why
While libfuzzer can be used as a library, engaging with it from some environments may be difficult to setup. Fazi provides similar functionality to libfuzzer, but gives greater flexibility into how you can use it. For instance, a native application which requires its own main entry point may be setup like:
/// compiled with -fsanitize=fuzer-no-link
extern "C" int LLVMFuzzerRunDriver(int *argc, char ***argv,
int (*UserCb)(const uint8_t *Data, size_t Size));
extern "C" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {
// fuzz
}
int main(int *argc, char ***argv) {
// Do my own thing
LLVMFuzzerRunDriver(argc, argv, LLVMFuzzerTestOneInput);
}
This model is difficult when integrating into an application with a lot of state or its own runtime environment, such as the JVM. Instead of providing a callback, Fazi lets you just ask it for data and tell it when the testcase is done:
extern "C" void fazi_start_iteration(char** data, size_t* size);
extern "C" void fazi_end_iteration(bool need_more_data);
extern "C" void fazi_initialize();
extern "C" void fazi_set_corpus_dir(const char*);
extern "C" void fazi_set_crashes_dir(const char*);
int main() {
// Setup fazi globals
fazi_initialize();
while (true) {
const char* data = nullptr;
size_t len = 0;
fazi_start_iteration(&data, &len);
bool need_more_data = some_api(data, len);
fazi_end_iteration(need_more_data);
}
}
Dependencies
~4MB
~82K SLoC