v8/test/cctest/wasm/test-c-wasm-entry.cc
Daniel Lehmann 8fffd56f86 [wasm] Allow execution while modifying code space
The --wasm-write-protect-code-memory flag previously enforced W^X, that
is the WebAssembly code space was either writable or executable, but
never both at the same time. With compilation in background threads
concurrent to execution in the main thread, this simple scheme is no
longer viable because the same memory page can indeed be written to and
executed at the same time. Hence, this flag is currently broken and
disabled and the code space is always writable AND executable.

As a first step towards more security, we at least want to
write-protect the code space (when not required writable by compilation
threads) but at the same time keep it always executable (because of
concurrent execution in the main thread). That is, we no longer switch
between RX and RW (W^X), but rather between RX and RWX
(write-protection only).

This CL starts to change from W^X (which was broken) to
write-protection only when enabling --wasm-write-protect-code-memory.
This is the first of two CLs, where the followup CL will fix the
feature, and this CL merely prepares and cleans up the code. In
particular, this CL changes the permissions from RW to RWX (due to
concurrent execution) and renames `WasmCodeAllocator::SetExecutable()`
to `WasmCodeAllocator::SetWritable()` (and similarly named callers) to
be consistent with that change. Since the code space is now always
executable, this CL also removes now unneeded calls to
`SetExecutable(true)` in tests.

R=clemensb@chromium.org
CC=​​jkummerow@chromium.org

Bug: v8:11663
Change-Id: I2065eed6770215892b81daefbddf74a349e783cc
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2835237
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Commit-Queue: Daniel Lehmann <dlehmann@google.com>
Cr-Commit-Position: refs/heads/master@{#74041}
2021-04-19 15:09:46 +00:00

188 lines
6.5 KiB
C++

// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include <cstdint>
#include "src/base/overflowing-math.h"
#include "src/base/safe_conversions.h"
#include "src/codegen/assembler-inl.h"
#include "src/objects/objects-inl.h"
#include "src/wasm/wasm-arguments.h"
#include "src/wasm/wasm-objects.h"
#include "test/cctest/cctest.h"
#include "test/cctest/compiler/value-helper.h"
#include "test/cctest/wasm/wasm-run-utils.h"
#include "test/common/wasm/wasm-macro-gen.h"
namespace v8 {
namespace internal {
namespace wasm {
/**
* We test the interface from C to compiled wasm code by generating a wasm
* function, creating a corresponding signature, compiling the c wasm entry for
* that signature, and then calling that entry using different test values.
* The result is compared against the expected result, computed from a lambda
* passed to the CWasmEntryArgTester.
*/
namespace {
template <typename ReturnType, typename... Args>
class CWasmEntryArgTester {
public:
CWasmEntryArgTester(std::initializer_list<uint8_t> wasm_function_bytes,
std::function<ReturnType(Args...)> expected_fn)
: runner_(TestExecutionTier::kTurbofan),
isolate_(runner_.main_isolate()),
expected_fn_(expected_fn),
sig_(runner_.template CreateSig<ReturnType, Args...>()) {
std::vector<uint8_t> code{wasm_function_bytes};
runner_.Build(code.data(), code.data() + code.size());
wasm_code_ = runner_.builder().GetFunctionCode(0);
c_wasm_entry_ = compiler::CompileCWasmEntry(
isolate_, sig_, wasm_code_->native_module()->module());
}
template <typename... Rest>
void WriteToBuffer(CWasmArgumentsPacker* packer, Rest... rest) {
static_assert(sizeof...(rest) == 0, "this is the base case");
}
template <typename First, typename... Rest>
void WriteToBuffer(CWasmArgumentsPacker* packer, First first, Rest... rest) {
packer->Push(first);
WriteToBuffer(packer, rest...);
}
void CheckCall(Args... args) {
CWasmArgumentsPacker packer(CWasmArgumentsPacker::TotalSize(sig_));
WriteToBuffer(&packer, args...);
Address wasm_call_target = wasm_code_->instruction_start();
Handle<Object> object_ref = runner_.builder().instance_object();
Execution::CallWasm(isolate_, c_wasm_entry_, wasm_call_target, object_ref,
packer.argv());
CHECK(!isolate_->has_pending_exception());
packer.Reset();
// Check the result.
ReturnType result = packer.Pop<ReturnType>();
ReturnType expected = expected_fn_(args...);
if (std::is_floating_point<ReturnType>::value) {
CHECK_DOUBLE_EQ(expected, result);
} else {
CHECK_EQ(expected, result);
}
}
private:
WasmRunner<ReturnType, Args...> runner_;
Isolate* isolate_;
std::function<ReturnType(Args...)> expected_fn_;
const FunctionSig* sig_;
Handle<Code> c_wasm_entry_;
WasmCode* wasm_code_;
};
} // namespace
// Pass int32_t, return int32_t.
TEST(TestCWasmEntryArgPassing_int32) {
CWasmEntryArgTester<int32_t, int32_t> tester(
{// Return 2*<0> + 1.
WASM_I32_ADD(WASM_I32_MUL(WASM_I32V_1(2), WASM_LOCAL_GET(0)), WASM_ONE)},
[](int32_t a) {
return base::AddWithWraparound(base::MulWithWraparound(2, a), 1);
});
FOR_INT32_INPUTS(v) { tester.CheckCall(v); }
}
// Pass int64_t, return double.
TEST(TestCWasmEntryArgPassing_double_int64) {
CWasmEntryArgTester<double, int64_t> tester(
{// Return (double)<0>.
WASM_F64_SCONVERT_I64(WASM_LOCAL_GET(0))},
[](int64_t a) { return static_cast<double>(a); });
FOR_INT64_INPUTS(v) { tester.CheckCall(v); }
}
// Pass double, return int64_t.
TEST(TestCWasmEntryArgPassing_int64_double) {
CWasmEntryArgTester<int64_t, double> tester(
{// Return (int64_t)<0>.
WASM_I64_SCONVERT_F64(WASM_LOCAL_GET(0))},
[](double d) { return static_cast<int64_t>(d); });
FOR_FLOAT64_INPUTS(d) {
if (base::IsValueInRangeForNumericType<int64_t>(d)) {
tester.CheckCall(d);
}
}
}
// Pass float, return double.
TEST(TestCWasmEntryArgPassing_float_double) {
CWasmEntryArgTester<double, float> tester(
{// Return 2*(double)<0> + 1.
WASM_F64_ADD(
WASM_F64_MUL(WASM_F64(2), WASM_F64_CONVERT_F32(WASM_LOCAL_GET(0))),
WASM_F64(1))},
[](float f) { return 2. * static_cast<double>(f) + 1.; });
FOR_FLOAT32_INPUTS(f) { tester.CheckCall(f); }
}
// Pass two doubles, return double.
TEST(TestCWasmEntryArgPassing_double_double) {
CWasmEntryArgTester<double, double, double> tester(
{// Return <0> + <1>.
WASM_F64_ADD(WASM_LOCAL_GET(0), WASM_LOCAL_GET(1))},
[](double a, double b) { return a + b; });
FOR_FLOAT64_INPUTS(d1) {
FOR_FLOAT64_INPUTS(d2) { tester.CheckCall(d1, d2); }
}
}
// Pass int32_t, int64_t, float and double, return double.
TEST(TestCWasmEntryArgPassing_AllTypes) {
CWasmEntryArgTester<double, int32_t, int64_t, float, double> tester(
{
// Convert all arguments to double, add them and return the sum.
WASM_F64_ADD( // <0+1+2> + <3>
WASM_F64_ADD( // <0+1> + <2>
WASM_F64_ADD( // <0> + <1>
WASM_F64_SCONVERT_I32(
WASM_LOCAL_GET(0)), // <0> to double
WASM_F64_SCONVERT_I64(
WASM_LOCAL_GET(1))), // <1> to double
WASM_F64_CONVERT_F32(WASM_LOCAL_GET(2))), // <2> to double
WASM_LOCAL_GET(3)) // <3>
},
[](int32_t a, int64_t b, float c, double d) {
return 0. + a + b + c + d;
});
Vector<const int32_t> test_values_i32 = compiler::ValueHelper::int32_vector();
Vector<const int64_t> test_values_i64 = compiler::ValueHelper::int64_vector();
Vector<const float> test_values_f32 = compiler::ValueHelper::float32_vector();
Vector<const double> test_values_f64 =
compiler::ValueHelper::float64_vector();
size_t max_len =
std::max(std::max(test_values_i32.size(), test_values_i64.size()),
std::max(test_values_f32.size(), test_values_f64.size()));
for (size_t i = 0; i < max_len; ++i) {
int32_t i32 = test_values_i32[i % test_values_i32.size()];
int64_t i64 = test_values_i64[i % test_values_i64.size()];
float f32 = test_values_f32[i % test_values_f32.size()];
double f64 = test_values_f64[i % test_values_f64.size()];
tester.CheckCall(i32, i64, f32, f64);
}
}
} // namespace wasm
} // namespace internal
} // namespace v8