[wasm] Add UTF-8 validation

Names passed for imports and exports are checked during decoding,
leading to errors if they are no valid UTF-8. Function names are not
checked during decode, but rather lead to undefined being returned at
runtime if they are not UTF-8.

We need to do these checks on the Wasm side, since the factory
methods assume to get valid UTF-8 strings.

R=titzer@chromium.org, yangguo@chromium.org

Review-Url: https://codereview.chromium.org/1967023004
Cr-Commit-Position: refs/heads/master@{#36208}
This commit is contained in:
clemensh 2016-05-12 06:01:10 -07:00 committed by Commit bot
parent 4aec7ba1aa
commit f0523e3046
8 changed files with 187 additions and 23 deletions

View File

@ -137,6 +137,12 @@ unsigned Utf8::Length(uchar c, int previous) {
}
}
bool Utf8::IsValidCharacter(uchar c) {
return c < 0xD800u || (c >= 0xE000u && c < 0xFDD0u) ||
(c > 0xFDEFu && c <= 0x10FFFFu && (c & 0xFFFEu) != 0xFFFEu &&
c != kBadChar);
}
} // namespace unibrow
#endif // V8_UNICODE_INL_H_

View File

@ -305,6 +305,20 @@ uchar Utf8::CalculateValue(const byte* str, size_t max_length, size_t* cursor) {
0x03C82080;
}
bool Utf8::Validate(const byte* bytes, size_t length) {
size_t cursor = 0;
// Performance optimization: Skip over single-byte values first.
while (cursor < length && bytes[cursor] <= kMaxOneByteChar) {
++cursor;
}
while (cursor < length) {
uchar c = ValueOf(bytes + cursor, length - cursor, &cursor);
if (!IsValidCharacter(c)) return false;
}
return true;
}
// Uppercase: point.category == 'Lu'

View File

@ -155,6 +155,11 @@ class Utf8 {
// UTF-8.
static const unsigned kMax16BitCodeUnitSize = 3;
static inline uchar ValueOf(const byte* str, size_t length, size_t* cursor);
// Excludes non-characters from the set of valid code points.
static inline bool IsValidCharacter(uchar c);
static bool Validate(const byte* str, size_t length);
};
struct Uppercase {

View File

@ -267,13 +267,13 @@ class ModuleDecoder : public Decoder {
for (uint32_t i = 0; i < functions_count; i++) {
WasmFunction* function = &module->functions[i];
function->name_offset =
consume_string(&function->name_length, "function name");
consume_string(&function->name_length, false);
uint32_t local_names_count =
consume_u32v(&length, "local names count");
for (uint32_t j = 0; j < local_names_count; j++) {
uint32_t unused = 0;
uint32_t offset = consume_string(&unused, "local name");
uint32_t offset = consume_string(&unused, false);
USE(unused);
USE(offset);
}
@ -372,13 +372,13 @@ class ModuleDecoder : public Decoder {
import->sig_index = consume_sig_index(module, &import->sig);
const byte* pos = pc_;
import->module_name_offset = consume_string(
&import->module_name_length, "import module name");
import->module_name_offset =
consume_string(&import->module_name_length, true);
if (import->module_name_length == 0) {
error(pos, "import module name cannot be NULL");
}
import->function_name_offset = consume_string(
&import->function_name_length, "import function name");
import->function_name_offset =
consume_string(&import->function_name_length, true);
}
break;
}
@ -402,7 +402,7 @@ class ModuleDecoder : public Decoder {
WasmFunction* func;
exp->func_index = consume_func_index(module, &func);
exp->name_offset = consume_string(&exp->name_length, "export name");
exp->name_offset = consume_string(&exp->name_length, true);
}
break;
}
@ -500,7 +500,9 @@ class ModuleDecoder : public Decoder {
// Decodes a single global entry inside a module starting at {pc_}.
void DecodeGlobalInModule(WasmGlobal* global) {
global->name_offset = consume_string(&global->name_length, "global name");
global->name_offset = consume_string(&global->name_length, false);
DCHECK(unibrow::Utf8::Validate(start_ + global->name_offset,
global->name_length));
global->type = mem_type();
global->offset = 0;
global->exported = consume_u8("exported") != 0;
@ -529,13 +531,13 @@ class ModuleDecoder : public Decoder {
decl_bits & kDeclFunctionExport ? " exported" : "",
(decl_bits & kDeclFunctionImport) == 0 ? " body" : "");
function->exported = decl_bits & kDeclFunctionExport;
if (decl_bits & kDeclFunctionName) {
function->name_offset =
consume_string(&function->name_length, "function name");
consume_string(&function->name_length, function->exported);
}
function->exported = decl_bits & kDeclFunctionExport;
// Imported functions have no locals or body.
if (decl_bits & kDeclFunctionImport) {
function->external = true;
@ -639,11 +641,14 @@ class ModuleDecoder : public Decoder {
// Reads a length-prefixed string, checking that it is within bounds. Returns
// the offset of the string, and the length as an out parameter.
uint32_t consume_string(uint32_t* length, const char* name = nullptr) {
uint32_t consume_string(uint32_t* length, bool validate_utf8) {
int varint_length;
*length = consume_u32v(&varint_length, "string length");
uint32_t offset = pc_offset();
TRACE(" +%u %-20s: (%u bytes)\n", offset, "string", *length);
if (validate_utf8 && !unibrow::Utf8::Validate(pc_, *length)) {
error(pc_, "no valid UTF-8 string");
}
consume_bytes(*length);
return offset;
}

View File

@ -60,6 +60,9 @@ Handle<Object> GetWasmFunctionNameFromTable(Handle<ByteArray> func_names_array,
: func_names_array->get_int(func_index + 2);
ScopedVector<byte> buffer(next_offset - offset);
func_names_array->copy_out(offset, buffer.start(), next_offset - offset);
if (!unibrow::Utf8::Validate(buffer.start(), buffer.length())) {
return undefined();
}
MaybeHandle<Object> maybe_name =
func_names_array->GetIsolate()->factory()->NewStringFromUtf8(
Vector<const char>::cast(buffer));

View File

@ -586,9 +586,9 @@ bool FinishCompilation(Isolate* isolate, WasmModule* module,
DCHECK_EQ(i, func.func_index);
WasmName str = module->GetName(func.name_offset, func.name_length);
WasmName str_null = {nullptr, 0};
Handle<String> name = factory->InternalizeUtf8String(str);
Handle<Code> code = Handle<Code>::null();
Handle<JSFunction> function = Handle<JSFunction>::null();
Handle<String> function_name = Handle<String>::null();
if (func.external) {
// Lookup external function in FFI object.
MaybeHandle<JSFunction> function =
@ -613,8 +613,9 @@ bool FinishCompilation(Isolate* isolate, WasmModule* module,
return false;
}
if (func.exported) {
function_name = factory->InternalizeUtf8String(str);
function = compiler::CompileJSToWasmWrapper(
isolate, &module_env, name, code, instance.js_object, i);
isolate, &module_env, function_name, code, instance.js_object, i);
record_code_size(total_code_size, function->code());
}
}
@ -629,7 +630,8 @@ bool FinishCompilation(Isolate* isolate, WasmModule* module,
// module.
desc.set_value(function);
Maybe<bool> status = JSReceiver::DefineOwnProperty(
isolate, instance.js_object, name, &desc, Object::THROW_ON_ERROR);
isolate, instance.js_object, function_name, &desc,
Object::THROW_ON_ERROR);
if (!status.IsJust())
thrower.Error("export of %.*s failed.", str.length(), str.start());
}

View File

@ -0,0 +1,121 @@
// Copyright 2016 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.
// Flags: --expose-wasm
load("test/mjsunit/wasm/wasm-constants.js");
load("test/mjsunit/wasm/wasm-module-builder.js");
function toByteArray(s) {
var arr = [];
for (var i = 0; i < s.length; ++i) {
arr.push(s.charCodeAt(i) & 0xff);
}
return arr;
}
function toString(arr) {
if (typeof arr === "string") return arr;
var s = "";
for (var b of arr) s += String.fromCharCode(b);
return s;
}
function toUTF8(arr) {
if (typeof arr === "string" || arr === undefined) return arr;
return decodeURIComponent(escape(toString(arr)));
}
function isValidUtf8(arr) {
if (typeof arr === "string" || arr === undefined) return true;
try {
var s = toUTF8(arr);
for (var i = 0; i < s.length; ++i)
if ((s.charCodeAt(i) & 0xfffe) == 0xfffe)
return false;
return true;
} catch (e) {
if (e instanceof URIError) return false;
throw e;
}
}
function checkImportsAndExports(imported_module_name, imported_function_name,
internal_function_name, exported_function_name, shouldThrow) {
var builder = new WasmModuleBuilder();
builder.addImportWithModule(imported_module_name, imported_function_name,
kSig_v_v);
builder.addFunction(internal_function_name, kSig_v_v)
.addBody([kExprCallImport, kArity0, 0])
.exportAs(exported_function_name);
// sanity check: does javascript agree with out shouldThrow annotation?
assertEquals(shouldThrow,
!isValidUtf8(imported_module_name) ||
!isValidUtf8(imported_function_name) ||
!isValidUtf8(exported_function_name),
"JavaScript does not agree with our shouldThrow expectation");
if (!shouldThrow) {
imported_module_name = toUTF8(imported_module_name);
imported_function_name = toUTF8(imported_function_name);
}
var ffi = new Object();
if (imported_function_name === undefined) {
ffi[imported_module_name] = function() { };
} else {
ffi[imported_module_name] = new Object();
ffi[imported_module_name][imported_function_name] = function() { };
}
var hasThrown = true;
try {
builder.instantiate(ffi);
hasThrown = false;
} catch (err) {
if (!shouldThrow) print(err);
assertTrue(shouldThrow, "Should not throw error on valid names");
assertContains("UTF-8", err.toString());
}
assertEquals(shouldThrow, hasThrown,
"Should throw validation error on invalid names");
}
function checkImportedModuleName(name, shouldThrow) {
checkImportsAndExports(name, "imp", "func", undefined, shouldThrow);
}
function checkImportedFunctionName(name, shouldThrow) {
checkImportsAndExports("module", name, "func", "func", shouldThrow);
}
function checkExportedFunctionName(name, shouldThrow) {
checkImportsAndExports("module", "func", "func", name, shouldThrow);
}
function checkInternalFunctionName(name) {
checkImportsAndExports("module", "func", name, "func", false);
}
function checkAll(name, shouldThrow) {
checkImportedModuleName(name, shouldThrow);
checkImportedFunctionName(name, shouldThrow);
checkExportedFunctionName(name, shouldThrow);
checkInternalFunctionName(name);
}
checkAll("ascii", false);
checkAll("some math: (½)² = ¼", false);
checkAll("中国历史系列条目\n北", false);
checkAll(toByteArray("\xef\xb7\x8f"), false);
checkAll(toByteArray("a\xc2\x81\xe1\x80\xbf\xf1\x80\xa0\xbf"), false);
checkAll(toByteArray("\xff"), true);
checkAll(toByteArray("\xed\xa0\x8f"), true); // surrogate code points
checkAll(toByteArray("\xe0\x82\x80"), true); // overlong sequence
checkAll(toByteArray("\xf4\x90\x80\x80"), true); // beyond limit: U+110000
checkAll(toByteArray("\xef\xbf\xbe"), true); // non-character; U+FFFE
checkAll(toByteArray("with\x00null"), false);

View File

@ -75,9 +75,7 @@ WasmModuleBuilder.prototype.addImportWithModule = function(module, name, sig) {
}
WasmModuleBuilder.prototype.addImport = function(name, sig) {
var sig_index = (typeof sig) == "number" ? sig : this.addSignature(sig);
this.imports.push({module: name, name: undefined, sig_index: sig_index});
return this.imports.length - 1;
this.addImportWithModule(name, undefined, sig);
}
WasmModuleBuilder.prototype.addDataSegment = function(addr, data, init) {
@ -107,9 +105,19 @@ function emit_u32(bytes, val) {
}
function emit_string(bytes, string) {
emit_varint(bytes, string.length);
for (var i = 0; i < string.length; i++) {
emit_u8(bytes, string.charCodeAt(i));
// When testing illegal names, we pass a byte array directly.
if (string instanceof Array) {
emit_varint(bytes, string.length);
emit_bytes(bytes, string);
return;
}
// This is the hacky way to convert a JavaScript scring to a UTF8 encoded
// string only containing single-byte characters.
var string_utf8 = unescape(encodeURIComponent(string));
emit_varint(bytes, string_utf8.length);
for (var i = 0; i < string_utf8.length; i++) {
emit_u8(bytes, string_utf8.charCodeAt(i));
}
}
@ -309,8 +317,8 @@ WasmModuleBuilder.prototype.toArray = function(debug) {
emit_varint(bytes, wasm.functions.length);
for (func of wasm.functions) {
var name = func.name == undefined ? "" : func.name;
emit_string(bytes, name);
emit_u8(bytes, 0); // local names count == 0
emit_string(bytes, name);
emit_u8(bytes, 0); // local names count == 0
}
});
}