* Add CLI flags for internal cranelift options This commit adds two flags to the `wasmtime` CLI: * `--enable-cranelift-debug-verifier` * `--enable-cranelift-nan-canonicalization` These previously weren't exposed from the command line but have been useful to me at least for reproducing slowdowns found during fuzzing on the CLI. * Disable Cranelift debug verifier when fuzzing This commit disables Cranelift's debug verifier for our fuzz targets. We've gotten a good number of timeouts on OSS-Fuzz and some I've recently had some discussion over at google/oss-fuzz#3944 about this issue and what we can do. The result of that discussion was that there are two primary ways we can speed up our fuzzers: * One is independent of Wasmtime, which is to tweak the flags used to compile code. The conclusion was that one flag was passed to LLVM which significantly increased runtime for very little benefit. This has now been disabled in rust-fuzz/cargo-fuzz#229. * The other way is to reduce the amount of debug checks we run while fuzzing wasmtime itself. To put this in perspective, a test case which took ~100ms to instantiate was taking 50 *seconds* to instantiate in the fuzz target. This 500x slowdown was caused by a ton of multiplicative factors, but two major contributors were NaN canonicalization and cranelift's debug verifier. I suspect the NaN canonicalization itself isn't too pricy but when paired with the debug verifier in float-heavy code it can create lots of IR to verify. This commit is specifically tackling this second point in an attempt to avoid slowing down our fuzzers too much. The intent here is that we'll disable the cranelift debug verifier for now but leave all other checks enabled. If the debug verifier gets a speed boost we can try re-enabling it, but otherwise it seems like for now it's otherwise not catching any bugs and creating lots of noise about timeouts that aren't relevant. It's not great that we have to turn off internal checks since that's what fuzzing is supposed to trigger, but given the timeout on OSS-Fuzz and the multiplicative effects of all the slowdowns we have when fuzzing, I'm not sure we can afford the massive slowdown of the debug verifier.
403 lines
15 KiB
Rust
403 lines
15 KiB
Rust
//! Oracles.
|
|
//!
|
|
//! Oracles take a test case and determine whether we have a bug. For example,
|
|
//! one of the simplest oracles is to take a Wasm binary as our input test case,
|
|
//! validate and instantiate it, and (implicitly) check that no assertions
|
|
//! failed or segfaults happened. A more complicated oracle might compare the
|
|
//! result of executing a Wasm file with and without optimizations enabled, and
|
|
//! make sure that the two executions are observably identical.
|
|
//!
|
|
//! When an oracle finds a bug, it should report it to the fuzzing engine by
|
|
//! panicking.
|
|
|
|
pub mod dummy;
|
|
|
|
use dummy::dummy_imports;
|
|
use std::sync::atomic::{AtomicUsize, Ordering::SeqCst};
|
|
use wasmtime::*;
|
|
use wasmtime_wast::WastContext;
|
|
|
|
fn log_wasm(wasm: &[u8]) {
|
|
static CNT: AtomicUsize = AtomicUsize::new(0);
|
|
if !log::log_enabled!(log::Level::Debug) {
|
|
return;
|
|
}
|
|
|
|
let i = CNT.fetch_add(1, SeqCst);
|
|
let name = format!("testcase{}.wasm", i);
|
|
std::fs::write(&name, wasm).expect("failed to write wasm file");
|
|
log::debug!("wrote wasm file to `{}`", name);
|
|
if let Ok(s) = wasmprinter::print_bytes(wasm) {
|
|
let name = format!("testcase{}.wat", i);
|
|
std::fs::write(&name, s).expect("failed to write wat file");
|
|
}
|
|
}
|
|
|
|
/// Instantiate the Wasm buffer, and implicitly fail if we have an unexpected
|
|
/// panic or segfault or anything else that can be detected "passively".
|
|
///
|
|
/// Performs initial validation, and returns early if the Wasm is invalid.
|
|
///
|
|
/// You can control which compiler is used via passing a `Strategy`.
|
|
pub fn instantiate(wasm: &[u8], strategy: Strategy) {
|
|
instantiate_with_config(wasm, crate::fuzz_default_config(strategy).unwrap());
|
|
}
|
|
|
|
/// Instantiate the Wasm buffer, and implicitly fail if we have an unexpected
|
|
/// panic or segfault or anything else that can be detected "passively".
|
|
///
|
|
/// The engine will be configured using provided config.
|
|
///
|
|
/// See also `instantiate` functions.
|
|
pub fn instantiate_with_config(wasm: &[u8], config: Config) {
|
|
crate::init_fuzzing();
|
|
|
|
let engine = Engine::new(&config);
|
|
let store = Store::new(&engine);
|
|
|
|
log_wasm(wasm);
|
|
let module = match Module::new(&engine, wasm) {
|
|
Ok(module) => module,
|
|
Err(_) => return,
|
|
};
|
|
|
|
let imports = match dummy_imports(&store, module.imports()) {
|
|
Ok(imps) => imps,
|
|
Err(_) => {
|
|
// There are some value types that we can't synthesize a
|
|
// dummy value for (e.g. externrefs) and for modules that
|
|
// import things of these types we skip instantiation.
|
|
return;
|
|
}
|
|
};
|
|
|
|
// Don't unwrap this: there can be instantiation-/link-time errors that
|
|
// aren't caught during validation or compilation. For example, an imported
|
|
// table might not have room for an element segment that we want to
|
|
// initialize into it.
|
|
let _result = Instance::new(&store, &module, &imports);
|
|
}
|
|
|
|
/// Compile the Wasm buffer, and implicitly fail if we have an unexpected
|
|
/// panic or segfault or anything else that can be detected "passively".
|
|
///
|
|
/// Performs initial validation, and returns early if the Wasm is invalid.
|
|
///
|
|
/// You can control which compiler is used via passing a `Strategy`.
|
|
pub fn compile(wasm: &[u8], strategy: Strategy) {
|
|
crate::init_fuzzing();
|
|
|
|
let engine = Engine::new(&crate::fuzz_default_config(strategy).unwrap());
|
|
log_wasm(wasm);
|
|
let _ = Module::new(&engine, wasm);
|
|
}
|
|
|
|
/// Instantiate the given Wasm module with each `Config` and call all of its
|
|
/// exports. Modulo OOM, non-canonical NaNs, and usage of Wasm features that are
|
|
/// or aren't enabled for different configs, we should get the same results when
|
|
/// we call the exported functions for all of our different configs.
|
|
#[cfg(feature = "binaryen")]
|
|
pub fn differential_execution(
|
|
ttf: &crate::generators::WasmOptTtf,
|
|
configs: &[crate::generators::DifferentialConfig],
|
|
) {
|
|
use std::collections::{HashMap, HashSet};
|
|
|
|
crate::init_fuzzing();
|
|
|
|
// We need at least two configs.
|
|
if configs.len() < 2
|
|
// And all the configs should be unique.
|
|
|| configs.iter().collect::<HashSet<_>>().len() != configs.len()
|
|
{
|
|
return;
|
|
}
|
|
|
|
let configs: Vec<_> = match configs.iter().map(|c| c.to_wasmtime_config()).collect() {
|
|
Ok(cs) => cs,
|
|
// If the config is trying to use something that was turned off at
|
|
// compile time, eg lightbeam, just continue to the next fuzz input.
|
|
Err(_) => return,
|
|
};
|
|
|
|
let mut export_func_results: HashMap<String, Result<Box<[Val]>, Trap>> = Default::default();
|
|
log_wasm(&ttf.wasm);
|
|
|
|
for config in &configs {
|
|
let engine = Engine::new(config);
|
|
let store = Store::new(&engine);
|
|
|
|
let module = match Module::new(&engine, &ttf.wasm) {
|
|
Ok(module) => module,
|
|
// The module might rely on some feature that our config didn't
|
|
// enable or something like that.
|
|
Err(e) => {
|
|
eprintln!("Warning: failed to compile `wasm-opt -ttf` module: {}", e);
|
|
continue;
|
|
}
|
|
};
|
|
|
|
// TODO: we should implement tracing versions of these dummy imports
|
|
// that record a trace of the order that imported functions were called
|
|
// in and with what values. Like the results of exported functions,
|
|
// calls to imports should also yield the same values for each
|
|
// configuration, and we should assert that.
|
|
let imports = match dummy_imports(&store, module.imports()) {
|
|
Ok(imps) => imps,
|
|
Err(e) => {
|
|
// There are some value types that we can't synthesize a
|
|
// dummy value for (e.g. externrefs) and for modules that
|
|
// import things of these types we skip instantiation.
|
|
eprintln!("Warning: failed to synthesize dummy imports: {}", e);
|
|
continue;
|
|
}
|
|
};
|
|
|
|
// Don't unwrap this: there can be instantiation-/link-time errors that
|
|
// aren't caught during validation or compilation. For example, an imported
|
|
// table might not have room for an element segment that we want to
|
|
// initialize into it.
|
|
let instance = match Instance::new(&store, &module, &imports) {
|
|
Ok(instance) => instance,
|
|
Err(e) => {
|
|
eprintln!(
|
|
"Warning: failed to instantiate `wasm-opt -ttf` module: {}",
|
|
e
|
|
);
|
|
continue;
|
|
}
|
|
};
|
|
|
|
for (name, f) in instance.exports().filter_map(|e| {
|
|
let name = e.name();
|
|
e.into_func().map(|f| (name, f))
|
|
}) {
|
|
// Always call the hang limit initializer first, so that we don't
|
|
// infinite loop when calling another export.
|
|
init_hang_limit(&instance);
|
|
|
|
let ty = f.ty();
|
|
let params = match dummy::dummy_values(ty.params()) {
|
|
Ok(p) => p,
|
|
Err(_) => continue,
|
|
};
|
|
let this_result = f.call(¶ms).map_err(|e| e.downcast::<Trap>().unwrap());
|
|
|
|
let existing_result = export_func_results
|
|
.entry(name.to_string())
|
|
.or_insert_with(|| this_result.clone());
|
|
assert_same_export_func_result(&existing_result, &this_result, name);
|
|
}
|
|
}
|
|
|
|
fn init_hang_limit(instance: &Instance) {
|
|
match instance.get_export("hangLimitInitializer") {
|
|
None => return,
|
|
Some(Extern::Func(f)) => {
|
|
f.call(&[])
|
|
.expect("initializing the hang limit should not fail");
|
|
}
|
|
Some(_) => panic!("unexpected hangLimitInitializer export"),
|
|
}
|
|
}
|
|
|
|
fn assert_same_export_func_result(
|
|
lhs: &Result<Box<[Val]>, Trap>,
|
|
rhs: &Result<Box<[Val]>, Trap>,
|
|
func_name: &str,
|
|
) {
|
|
let fail = || {
|
|
panic!(
|
|
"differential fuzzing failed: exported func {} returned two \
|
|
different results: {:?} != {:?}",
|
|
func_name, lhs, rhs
|
|
)
|
|
};
|
|
|
|
match (lhs, rhs) {
|
|
(Err(_), Err(_)) => {}
|
|
(Ok(lhs), Ok(rhs)) => {
|
|
if lhs.len() != rhs.len() {
|
|
fail();
|
|
}
|
|
for (lhs, rhs) in lhs.iter().zip(rhs.iter()) {
|
|
match (lhs, rhs) {
|
|
(Val::I32(lhs), Val::I32(rhs)) if lhs == rhs => continue,
|
|
(Val::I64(lhs), Val::I64(rhs)) if lhs == rhs => continue,
|
|
(Val::V128(lhs), Val::V128(rhs)) if lhs == rhs => continue,
|
|
(Val::F32(lhs), Val::F32(rhs)) => {
|
|
let lhs = f32::from_bits(*lhs);
|
|
let rhs = f32::from_bits(*rhs);
|
|
if lhs == rhs || (lhs.is_nan() && rhs.is_nan()) {
|
|
continue;
|
|
} else {
|
|
fail()
|
|
}
|
|
}
|
|
(Val::F64(lhs), Val::F64(rhs)) => {
|
|
let lhs = f64::from_bits(*lhs);
|
|
let rhs = f64::from_bits(*rhs);
|
|
if lhs == rhs || (lhs.is_nan() && rhs.is_nan()) {
|
|
continue;
|
|
} else {
|
|
fail()
|
|
}
|
|
}
|
|
(Val::ExternRef(_), Val::ExternRef(_))
|
|
| (Val::FuncRef(_), Val::FuncRef(_)) => continue,
|
|
_ => fail(),
|
|
}
|
|
}
|
|
}
|
|
_ => fail(),
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Invoke the given API calls.
|
|
#[cfg(feature = "binaryen")]
|
|
pub fn make_api_calls(api: crate::generators::api::ApiCalls) {
|
|
use crate::generators::api::ApiCall;
|
|
use std::collections::HashMap;
|
|
|
|
crate::init_fuzzing();
|
|
|
|
let mut config: Option<Config> = None;
|
|
let mut engine: Option<Engine> = None;
|
|
let mut store: Option<Store> = None;
|
|
let mut modules: HashMap<usize, Module> = Default::default();
|
|
let mut instances: HashMap<usize, Instance> = Default::default();
|
|
|
|
for call in api.calls {
|
|
match call {
|
|
ApiCall::ConfigNew => {
|
|
log::trace!("creating config");
|
|
assert!(config.is_none());
|
|
config = Some(crate::fuzz_default_config(wasmtime::Strategy::Cranelift).unwrap());
|
|
}
|
|
|
|
ApiCall::ConfigDebugInfo(b) => {
|
|
log::trace!("enabling debuginfo");
|
|
config.as_mut().unwrap().debug_info(b);
|
|
}
|
|
|
|
ApiCall::ConfigInterruptable(b) => {
|
|
log::trace!("enabling interruption");
|
|
config.as_mut().unwrap().interruptable(b);
|
|
}
|
|
|
|
ApiCall::EngineNew => {
|
|
log::trace!("creating engine");
|
|
assert!(engine.is_none());
|
|
engine = Some(Engine::new(config.as_ref().unwrap()));
|
|
}
|
|
|
|
ApiCall::StoreNew => {
|
|
log::trace!("creating store");
|
|
assert!(store.is_none());
|
|
store = Some(Store::new(engine.as_ref().unwrap()));
|
|
}
|
|
|
|
ApiCall::ModuleNew { id, wasm } => {
|
|
log::debug!("creating module: {}", id);
|
|
log_wasm(&wasm.wasm);
|
|
let module = match Module::new(engine.as_ref().unwrap(), &wasm.wasm) {
|
|
Ok(m) => m,
|
|
Err(_) => continue,
|
|
};
|
|
let old = modules.insert(id, module);
|
|
assert!(old.is_none());
|
|
}
|
|
|
|
ApiCall::ModuleDrop { id } => {
|
|
log::trace!("dropping module: {}", id);
|
|
drop(modules.remove(&id));
|
|
}
|
|
|
|
ApiCall::InstanceNew { id, module } => {
|
|
log::trace!("instantiating module {} as {}", module, id);
|
|
let module = match modules.get(&module) {
|
|
Some(m) => m,
|
|
None => continue,
|
|
};
|
|
|
|
let store = store.as_ref().unwrap();
|
|
|
|
let imports = match dummy_imports(store, module.imports()) {
|
|
Ok(imps) => imps,
|
|
Err(_) => {
|
|
// There are some value types that we can't synthesize a
|
|
// dummy value for (e.g. externrefs) and for modules that
|
|
// import things of these types we skip instantiation.
|
|
continue;
|
|
}
|
|
};
|
|
|
|
// Don't unwrap this: there can be instantiation-/link-time errors that
|
|
// aren't caught during validation or compilation. For example, an imported
|
|
// table might not have room for an element segment that we want to
|
|
// initialize into it.
|
|
if let Ok(instance) = Instance::new(store, &module, &imports) {
|
|
instances.insert(id, instance);
|
|
}
|
|
}
|
|
|
|
ApiCall::InstanceDrop { id } => {
|
|
log::trace!("dropping instance {}", id);
|
|
drop(instances.remove(&id));
|
|
}
|
|
|
|
ApiCall::CallExportedFunc { instance, nth } => {
|
|
log::trace!("calling instance export {} / {}", instance, nth);
|
|
let instance = match instances.get(&instance) {
|
|
Some(i) => i,
|
|
None => {
|
|
// Note that we aren't guaranteed to instantiate valid
|
|
// modules, see comments in `InstanceNew` for details on
|
|
// that. But the API call generator can't know if
|
|
// instantiation failed, so we might not actually have
|
|
// this instance. When that's the case, just skip the
|
|
// API call and keep going.
|
|
continue;
|
|
}
|
|
};
|
|
|
|
let funcs = instance
|
|
.exports()
|
|
.filter_map(|e| match e.into_extern() {
|
|
Extern::Func(f) => Some(f.clone()),
|
|
_ => None,
|
|
})
|
|
.collect::<Vec<_>>();
|
|
|
|
if funcs.is_empty() {
|
|
continue;
|
|
}
|
|
|
|
let nth = nth % funcs.len();
|
|
let f = &funcs[nth];
|
|
let ty = f.ty();
|
|
let params = match dummy::dummy_values(ty.params()) {
|
|
Ok(p) => p,
|
|
Err(_) => continue,
|
|
};
|
|
let _ = f.call(¶ms);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Executes the wast `test` spectest with the `config` specified.
|
|
///
|
|
/// Ensures that spec tests pass regardless of the `Config`.
|
|
pub fn spectest(config: crate::generators::Config, test: crate::generators::SpecTest) {
|
|
crate::init_fuzzing();
|
|
log::debug!("running {:?} with {:?}", test.file, config);
|
|
let store = Store::new(&Engine::new(&config.to_wasmtime()));
|
|
let mut wast_context = WastContext::new(store);
|
|
wast_context.register_spectest().unwrap();
|
|
wast_context
|
|
.run_buffer(test.file, test.contents.as_bytes())
|
|
.unwrap();
|
|
}
|