Files
wasmtime/crates/fuzzing/src/oracles.rs
Alex Crichton c9a0ba81a0 Implement interrupting wasm code, reimplement stack overflow (#1490)
* Implement interrupting wasm code, reimplement stack overflow

This commit is a relatively large change for wasmtime with two main
goals:

* Primarily this enables interrupting executing wasm code with a trap,
  preventing infinite loops in wasm code. Note that resumption of the
  wasm code is not a goal of this commit.

* Additionally this commit reimplements how we handle stack overflow to
  ensure that host functions always have a reasonable amount of stack to
  run on. This fixes an issue where we might longjmp out of a host
  function, skipping destructors.

Lots of various odds and ends end up falling out in this commit once the
two goals above were implemented. The strategy for implementing this was
also lifted from Spidermonkey and existing functionality inside of
Cranelift. I've tried to write up thorough documentation of how this all
works in `crates/environ/src/cranelift.rs` where gnarly-ish bits are.

A brief summary of how this works is that each function and each loop
header now checks to see if they're interrupted. Interrupts and the
stack overflow check are actually folded into one now, where function
headers check to see if they've run out of stack and the sentinel value
used to indicate an interrupt, checked in loop headers, tricks functions
into thinking they're out of stack. An interrupt is basically just
writing a value to a location which is read by JIT code.

When interrupts are delivered and what triggers them has been left up to
embedders of the `wasmtime` crate. The `wasmtime::Store` type has a
method to acquire an `InterruptHandle`, where `InterruptHandle` is a
`Send` and `Sync` type which can travel to other threads (or perhaps
even a signal handler) to get notified from. It's intended that this
provides a good degree of flexibility when interrupting wasm code. Note
though that this does have a large caveat where interrupts don't work
when you're interrupting host code, so if you've got a host import
blocking for a long time an interrupt won't actually be received until
the wasm starts running again.

Some fallout included from this change is:

* Unix signal handlers are no longer registered with `SA_ONSTACK`.
  Instead they run on the native stack the thread was already using.
  This is possible since stack overflow isn't handled by hitting the
  guard page, but rather it's explicitly checked for in wasm now. Native
  stack overflow will continue to abort the process as usual.

* Unix sigaltstack management is now no longer necessary since we don't
  use it any more.

* Windows no longer has any need to reset guard pages since we no longer
  try to recover from faults on guard pages.

* On all targets probestack intrinsics are disabled since we use a
  different mechanism for catching stack overflow.

* The C API has been updated with interrupts handles. An example has
  also been added which shows off how to interrupt a module.

Closes #139
Closes #860
Closes #900

* Update comment about magical interrupt value

* Store stack limit as a global value, not a closure

* Run rustfmt

* Handle review comments

* Add a comment about SA_ONSTACK

* Use `usize` for type of `INTERRUPTED`

* Parse human-readable durations

* Bring back sigaltstack handling

Allows libstd to print out stack overflow on failure still.

* Add parsing and emission of stack limit-via-preamble

* Fix new example for new apis

* Fix host segfault test in release mode

* Fix new doc example
2020-04-21 11:03:28 -07:00

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(&store, 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. anyrefs) 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(&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());
let store = Store::new(&engine);
log_wasm(wasm);
let _ = Module::new(&store, 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(&store, &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. anyrefs) 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(&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(&params).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::AnyRef(_), Val::AnyRef(_)) | (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());
let mut cfg = Config::new();
cfg.cranelift_debug_verifier(true);
config = Some(cfg);
}
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(store.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 imports = match dummy_imports(store.as_ref().unwrap(), module.imports()) {
Ok(imps) => imps,
Err(_) => {
// There are some value types that we can't synthesize a
// dummy value for (e.g. anyrefs) 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(&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(&params);
}
}
}
}
/// 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) {
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();
}