This directory contains the fuzz tests for wasmer. To fuzz, we use the cargo-fuzz
package.
You may need to install the cargo-fuzz
package to get the cargo fuzz
subcommand. Use
$ cargo install cargo-fuzz
cargo-fuzz
is documented in the Rust Fuzz Book.
Once cargo-fuzz
is installed, you can run the simple_instantiate
fuzzer with
cargo fuzz run simple_instantiate
or the validate_wasm
fuzzer
cargo fuzz run validate_wasm
or the compile_wasm
fuzzer
cargo fuzz run compile_wasm
You should see output that looks something like this:
INFO: Seed: 3276026494
INFO: 8 files found in wasmer/fuzz/corpus/simple_instantiate
INFO: -max_len is not provided; libFuzzer will not generate inputs larger than 4096 bytes
INFO: seed corpus: files: 8 min: 1b max: 1b total: 8b rss: 133Mb
#9 INITED ft: 3 corp: 3/3b lim: 4 exec/s: 0 rss: 142Mb
#23 NEW ft: 4 corp: 4/5b lim: 4 exec/s: 0 rss: 142Mb L: 2/2 MS: 4 ChangeByte-InsertByte-ShuffleBytes-ChangeBit-
#25 NEW ft: 5 corp: 5/6b lim: 4 exec/s: 0 rss: 142Mb L: 1/2 MS: 2 ChangeBinInt-ChangeBit-
#27 NEW ft: 6 corp: 6/9b lim: 4 exec/s: 0 rss: 142Mb L: 3/3 MS: 2 InsertByte-ChangeByte-
#190 REDUCE ft: 6 corp: 6/7b lim: 4 exec/s: 0 rss: 142Mb L: 1/2 MS: 3 ChangeBit-EraseBytes-CrossOver-
#205 REDUCE ft: 7 corp: 7/11b lim: 4 exec/s: 0 rss: 142Mb L: 4/4 MS: 5 ShuffleBytes-CrossOver-InsertByte-ChangeBinInt-CrossOver-
It will continue to generate random inputs forever, until it finds a bug or is terminated. The testcases for bugs it finds go into fuzz/artifacts/simple_instantiate
and you can rerun the fuzzer on a single input by passing it on the command line cargo fuzz run simple_instantiate my_testcase.wasm
.
The fuzzer works best when it has examples of small Wasm files to start with. Using wast2json
from wabt, we can easily produce .wasm
files out of the WebAssembly spec tests.
mkdir spec-test-corpus
for i in lib/spectests/spectests/*.wast; do wast2json --enable-all $i -o spec-test-corpus/$(basename $i).json; done
mv spec-test-corpus/*.wasm fuzz/corpus/simple_instantiate/
rm -r spec-test-corpus
The corpus directory is created on the first run of the fuzzer. If it doesn't exist, run it first and then seed the corpus. The fuzzer will pick up new files added to the corpus while it is running.