Home | History | Annotate | Line # | Download | only in fuzz
      1 # Fuzzing
      2 
      3 Each fuzzing target can be built with multiple engines.
      4 Zstd provides a fuzz corpus for each target that can be downloaded with
      5 the command:
      6 
      7 ```
      8 make corpora
      9 ```
     10 
     11 It will download each corpus into `./corpora/TARGET`.
     12 
     13 ## fuzz.py
     14 
     15 `fuzz.py` is a helper script for building and running fuzzers.
     16 Run `./fuzz.py -h` for the commands and run `./fuzz.py COMMAND -h` for
     17 command specific help.
     18 
     19 ### Generating Data
     20 
     21 `fuzz.py` provides a utility to generate seed data for each fuzzer.
     22 
     23 ```
     24 make -C ../tests decodecorpus
     25 ./fuzz.py gen TARGET
     26 ```
     27 
     28 By default it outputs 100 samples, each at most 8KB into `corpora/TARGET-seed`,
     29 but that can be configured with the `--number`, `--max-size-log` and `--seed`
     30 flags.
     31 
     32 ### Build
     33 It respects the usual build environment variables `CC`, `CFLAGS`, etc.
     34 The environment variables can be overridden with the corresponding flags
     35 `--cc`, `--cflags`, etc.
     36 The specific fuzzing engine is selected with `LIB_FUZZING_ENGINE` or
     37 `--lib-fuzzing-engine`, the default is `libregression.a`.
     38 Alternatively, you can use Clang's built in fuzzing engine with
     39 `--enable-fuzzer`.
     40 It has flags that can easily set up sanitizers `--enable-{a,ub,m}san`, and
     41 coverage instrumentation `--enable-coverage`.
     42 It sets sane defaults which can be overridden with flags `--debug`,
     43 `--enable-ubsan-pointer-overflow`, etc.
     44 Run `./fuzz.py build -h` for help.
     45 
     46 ### Running Fuzzers
     47 
     48 `./fuzz.py` can run `libfuzzer`, `afl`, and `regression` tests.
     49 See the help of the relevant command for options.
     50 Flags not parsed by `fuzz.py` are passed to the fuzzing engine.
     51 The command used to run the fuzzer is printed for debugging.
     52 
     53 Here's a helpful command to fuzz each target across all cores,
     54 stopping only if a bug is found:
     55 ```
     56 for target in $(./fuzz.py list); do
     57     ./fuzz.py libfuzzer $target -jobs=10 -workers=10 -max_total_time=1000 || break;
     58 done
     59 ```
     60 Alternatively, you can fuzz all targets in parallel, using one core per target:
     61 ```
     62 python3 ./fuzz.py list | xargs -P$(python3 ./fuzz.py list | wc -l) -I__ sh -c "python3 ./fuzz.py libfuzzer __ 2>&1 | tee __.log"
     63 ```
     64 Either way, to double-check that no crashes were found, run `ls corpora/*crash`.
     65 If any crashes were found, you can use the hashes to reproduce them.
     66 
     67 ## LibFuzzer
     68 
     69 ```
     70 # Build the fuzz targets
     71 ./fuzz.py build all --enable-fuzzer --enable-asan --enable-ubsan --cc clang --cxx clang++
     72 # OR equivalently
     73 CC=clang CXX=clang++ ./fuzz.py build all --enable-fuzzer --enable-asan --enable-ubsan
     74 # Run the fuzzer
     75 ./fuzz.py libfuzzer TARGET <libfuzzer args like -jobs=4>
     76 ```
     77 
     78 where `TARGET` could be `simple_decompress`, `stream_round_trip`, etc.
     79 
     80 ### MSAN
     81 
     82 Fuzzing with `libFuzzer` and `MSAN` is as easy as:
     83 
     84 ```
     85 CC=clang CXX=clang++ ./fuzz.py build all --enable-fuzzer --enable-msan
     86 ./fuzz.py libfuzzer TARGET <libfuzzer args>
     87 ```
     88 
     89 `fuzz.py` respects the environment variables / flags `MSAN_EXTRA_CPPFLAGS`,
     90 `MSAN_EXTRA_CFLAGS`, `MSAN_EXTRA_CXXFLAGS`, `MSAN_EXTRA_LDFLAGS` to easily pass
     91 the extra parameters only for MSAN.
     92 
     93 ## AFL
     94 
     95 The default `LIB_FUZZING_ENGINE` is `libregression.a`, which produces a binary
     96 that AFL can use.
     97 
     98 ```
     99 # Build the fuzz targets
    100 CC=afl-clang CXX=afl-clang++ ./fuzz.py build all --enable-asan --enable-ubsan
    101 # Run the fuzzer without a memory limit because of ASAN
    102 ./fuzz.py afl TARGET -m none
    103 ```
    104 
    105 ## Regression Testing
    106 
    107 The regression test supports the `all` target to run all the fuzzers in one
    108 command.
    109 
    110 ```
    111 CC=clang CXX=clang++ ./fuzz.py build all --enable-asan --enable-ubsan
    112 ./fuzz.py regression all
    113 CC=clang CXX=clang++ ./fuzz.py build all --enable-msan
    114 ./fuzz.py regression all
    115 ```
    116 
    117 ## Fuzzing a custom sequence producer plugin
    118 Sequence producer plugin authors can use the zstd fuzzers to stress-test their code.
    119 See the documentation in `fuzz_third_party_seq_prod.h` for details.
    120 
    121 ## Adding a new fuzzer
    122 There are several steps involved in adding a new fuzzer harness.
    123 
    124 ### Build your harness
    125 1. Create a new your fuzzer harness `tests/fuzz/your_harness.c`.
    126 
    127 2. Add your harness to the Makefile
    128 
    129     2.1 Follow [this example](https://github.com/facebook/zstd/blob/e124e39301381de8f323436a3e4c46539747ba24/tests/fuzz/Makefile#L216) if your fuzzer requires both compression and decompression symbols (prefix `rt_`). If your fuzzer only requires decompression symbols, follow [this example](https://github.com/facebook/zstd/blob/6a0052a409e2604bd40354b76b86272b712edd7d/tests/fuzz/Makefile#L194) (prefix `d_`).
    130     
    131     2.2 Add your target to [`FUZZ_TARGETS`](https://github.com/facebook/zstd/blob/6a0052a409e2604bd40354b76b86272b712edd7d/tests/fuzz/Makefile#L108).
    132     
    133 3. Add your harness to [`fuzz.py`](https://github.com/facebook/zstd/blob/6a0052a409e2604bd40354b76b86272b712edd7d/tests/fuzz/fuzz.py#L48).
    134 
    135 ### Generate seed data
    136 Follow the instructions above to generate seed data:
    137 ```
    138 make -C ../tests decodecorpus
    139 ./fuzz.py gen your_harness
    140 ```
    141 
    142 ### Run the harness
    143 Follow the instructions above to run your harness and fix any crashes:
    144 ```
    145 ./fuzz.py build your_harness --enable-fuzzer --enable-asan --enable-ubsan --cc clang --cxx clang++
    146 ./fuzz.py libfuzzer your_harness
    147 ```
    148 
    149 ### Minimize and zip the corpus
    150 After running the fuzzer for a while, you will have a large corpus at `tests/fuzz/corpora/your_harness*`.
    151 This corpus must be minimized and zipped before uploading to GitHub for regression testing:
    152 ```
    153 ./fuzz.py minimize your_harness
    154 ./fuzz.py zip your_harness 
    155 ```
    156 
    157 ### Upload the zip file to GitHub
    158 The previous step should produce a `.zip` file containing the corpus for your new harness.
    159 This corpus must be uploaded to GitHub here: https://github.com/facebook/zstd/releases/tag/fuzz-corpora
    160 
    161 
    162