void HTMLDocumentParser::constructTreeFromHTMLToken(HTMLToken& rawToken)
{
RefPtr<AtomicHTMLToken> token = AtomicHTMLToken::create(rawToken);
// We clear the rawToken in case constructTreeFromAtomicToken
// synchronously re-enters the parser. We don't clear the token immedately
// for Character tokens because the AtomicHTMLToken avoids copying the
// characters by keeping a pointer to the underlying buffer in the
// HTMLToken. Fortunately, Character tokens can't cause us to re-enter
// the parser.
//
// FIXME: Stop clearing the rawToken once we start running the parser off
// the main thread or once we stop allowing synchronous JavaScript
// execution from parseAttribute.
if (rawToken.type() != HTMLTokenTypes::Character)
rawToken.clear();
m_treeBuilder->constructTree(token.get());
// AtomicHTMLToken keeps a pointer to the HTMLToken's buffer instead
// of copying the characters for performance.
// Clear the external characters pointer before the raw token is cleared
// to make sure that we won't have a dangling pointer.
token->clearExternalCharacters();
if (!rawToken.isUninitialized()) {
ASSERT(rawToken.type() == HTMLTokenTypes::Character);
rawToken.clear();
}
}
void XSSAuditor::filterToken(HTMLToken& token)
{
if (m_state == Uninitialized)
init();
ASSERT(m_state == Initialized);
if (!m_isEnabled || m_xssProtection == XSSProtectionDisabled)
return;
bool didBlockScript = false;
if (token.type() == HTMLTokenTypes::StartTag)
didBlockScript = filterStartToken(token);
else if (m_scriptTagNestingLevel) {
if (token.type() == HTMLTokenTypes::Character)
didBlockScript = filterCharacterToken(token);
else if (token.type() == HTMLTokenTypes::EndTag)
filterEndToken(token);
}
if (didBlockScript) {
// FIXME: Consider using a more helpful console message.
DEFINE_STATIC_LOCAL(String, consoleMessage, (ASCIILiteral("Refused to execute a JavaScript script. Source code of script found within request.\n")));
m_parser->document()->addConsoleMessage(JSMessageSource, LogMessageType, ErrorMessageLevel, consoleMessage);
bool didBlockEntirePage = (m_xssProtection == XSSProtectionBlockEnabled);
if (didBlockEntirePage)
m_parser->document()->frame()->loader()->stopAllLoaders();
if (m_notifyClient) {
m_parser->document()->frame()->loader()->client()->didDetectXSS(m_parser->document()->url(), didBlockEntirePage);
m_notifyClient = false;
}
if (!m_reportURL.isEmpty()) {
RefPtr<InspectorObject> reportDetails = InspectorObject::create();
reportDetails->setString("request-url", m_originalURL);
reportDetails->setString("request-body", m_originalHTTPBody);
RefPtr<InspectorObject> reportObject = InspectorObject::create();
reportObject->setObject("xss-report", reportDetails.release());
RefPtr<FormData> report = FormData::create(reportObject->toJSONString().utf8().data());
PingLoader::sendViolationReport(m_parser->document()->frame(), m_reportURL, report);
m_reportURL = KURL();
m_originalURL = String();
m_originalHTTPBody = String();
}
if (didBlockEntirePage)
m_parser->document()->frame()->navigationScheduler()->scheduleLocationChange(m_parser->document()->securityOrigin(), blankURL(), String());
}
}
CompactHTMLToken::CompactHTMLToken(const HTMLToken& token)
: m_type(token.type())
{
switch (m_type) {
case HTMLTokenTypes::Uninitialized:
ASSERT_NOT_REACHED();
break;
case HTMLTokenTypes::DOCTYPE:
m_data = String(token.name().data(), token.name().size());
m_publicIdentifier = String(token.publicIdentifier().data(), token.publicIdentifier().size());
m_systemIdentifier = String(token.systemIdentifier().data(), token.systemIdentifier().size());
break;
case HTMLTokenTypes::EndOfFile:
break;
case HTMLTokenTypes::StartTag:
m_attributes.reserveInitialCapacity(token.attributes().size());
for (Vector<AttributeBase>::const_iterator it = token.attributes().begin(); it != token.attributes().end(); ++it)
m_attributes.append(CompactAttribute(String(it->m_name.data(), it->m_name.size()), String(it->m_value.data(), it->m_value.size())));
// Fall through!
case HTMLTokenTypes::EndTag:
m_selfClosing = token.selfClosing();
// Fall through!
case HTMLTokenTypes::Comment:
case HTMLTokenTypes::Character:
if (token.isAll8BitData())
m_data = String::make8BitFrom16BitSource(token.data().data(), token.data().size());
else
m_data = String(token.data().data(), token.data().size());
break;
default:
ASSERT_NOT_REACHED();
break;
}
}
void HTMLViewSourceDocument::addSource(const String& source, HTMLToken& token)
{
if (!m_current)
createContainingTable();
switch (token.type()) {
case HTMLToken::Uninitialized:
ASSERT_NOT_REACHED();
break;
case HTMLToken::DOCTYPE:
processDoctypeToken(source, token);
break;
case HTMLToken::EndOfFile:
break;
case HTMLToken::StartTag:
case HTMLToken::EndTag:
processTagToken(source, token);
break;
case HTMLToken::Comment:
processCommentToken(source, token);
break;
case HTMLToken::Character:
processCharacterToken(source, token);
break;
}
}
String HTMLSourceTracker::sourceForToken(const HTMLToken& token)
{
if (token.type() == HTMLToken::EndOfFile)
return String(); // Hides the null character we use to mark the end of file.
if (!m_cachedSourceForToken.isEmpty())
return m_cachedSourceForToken;
ASSERT(!token.startIndex());
size_t length = static_cast<size_t>(token.endIndex() - token.startIndex());
StringBuilder source;
source.reserveCapacity(length);
size_t i = 0;
for ( ; i < length && !m_previousSource.isEmpty(); ++i) {
source.append(m_previousSource.currentChar());
m_previousSource.advance();
}
for ( ; i < length; ++i) {
ASSERT(!m_currentSource.isEmpty());
source.append(m_currentSource.currentChar());
m_currentSource.advance();
}
m_cachedSourceForToken = source.toString();
return m_cachedSourceForToken;
}
String HTMLSourceTracker::sourceForToken(const HTMLToken& token)
{
if (!m_cachedSourceForToken.isEmpty())
return m_cachedSourceForToken;
size_t length;
if (token.type() == HTMLToken::EndOfFile) {
// Consume the remainder of the input, omitting the null character we use to mark the end of the file.
length = m_previousSource.length() + m_currentSource.length() - 1;
} else {
ASSERT(!token.startIndex());
length = static_cast<size_t>(token.endIndex() - token.startIndex());
}
StringBuilder source;
source.reserveCapacity(length);
size_t i = 0;
for ( ; i < length && !m_previousSource.isEmpty(); ++i) {
source.append(m_previousSource.currentChar());
m_previousSource.advance();
}
for ( ; i < length; ++i) {
ASSERT(!m_currentSource.isEmpty());
source.append(m_currentSource.currentChar());
m_currentSource.advance();
}
m_cachedSourceForToken = source.toString();
return m_cachedSourceForToken;
}
bool XSSAuditor::filterFormToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, formTag));
return eraseAttributeIfInjected(token, actionAttr, blankURL().string());
}
bool XSSAuditor::filterTokenInitial(HTMLToken& token)
{
ASSERT(m_state == Initial);
if (token.type() != HTMLTokenTypes::StartTag)
return false;
bool didBlockScript = eraseDangerousAttributesIfInjected(token);
if (hasName(token, scriptTag))
didBlockScript |= filterScriptToken(token);
else if (hasName(token, objectTag))
didBlockScript |= filterObjectToken(token);
else if (hasName(token, paramTag))
didBlockScript |= filterParamToken(token);
else if (hasName(token, embedTag))
didBlockScript |= filterEmbedToken(token);
else if (hasName(token, appletTag))
didBlockScript |= filterAppletToken(token);
else if (hasName(token, iframeTag))
didBlockScript |= filterIframeToken(token);
else if (hasName(token, metaTag))
didBlockScript |= filterMetaToken(token);
else if (hasName(token, baseTag))
didBlockScript |= filterBaseToken(token);
else if (hasName(token, formTag))
didBlockScript |= filterFormToken(token);
return didBlockScript;
}
bool XSSAuditor::filterBaseToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, baseTag));
return eraseAttributeIfInjected(token, hrefAttr);
}
bool XSSAuditor::filterMetaToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, metaTag));
return eraseAttributeIfInjected(token, http_equivAttr);
}
bool XSSFilter::filterBaseToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLToken::StartTag);
ASSERT(hasName(token, baseTag));
return eraseAttributeIfInjected(token, hrefAttr);
}
bool XSSFilter::filterMetaToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLToken::StartTag);
ASSERT(hasName(token, metaTag));
return eraseAttributeIfInjected(token, http_equivAttr);
}
bool XSSAuditor::filterIframeToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, iframeTag));
return eraseAttributeIfInjected(token, srcAttr, String(), SrcLikeAttribute);
}
bool XSSAuditor::filterFormToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, formTag));
return eraseAttributeIfInjected(token, actionAttr);
}
bool XSSFilter::filterTokenAfterScriptStartTag(HTMLToken& token)
{
ASSERT(m_state == AfterScriptStartTag);
m_state = Initial;
if (token.type() != HTMLToken::Character) {
ASSERT(token.type() == HTMLToken::EndTag || token.type() == HTMLToken::EndOfFile);
return false;
}
int start = 0;
// FIXME: We probably want to grab only the first few characters of the
// contents of the script element.
int end = token.endIndex() - token.startIndex();
if (isContainedInRequest(m_cachedSnippet + snippetForRange(token, start, end))) {
token.eraseCharacters();
token.appendToCharacter(' '); // Technically, character tokens can't be empty.
return true;
}
return false;
}
bool XSSAuditor::filterIframeToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, iframeTag));
bool didBlockScript = false;
if (isContainedInRequest(decodedSnippetForName(token))) {
didBlockScript |= eraseAttributeIfInjected(token, srcAttr, String(), SrcLikeAttribute);
didBlockScript |= eraseAttributeIfInjected(token, srcdocAttr, String(), ScriptLikeAttribute);
}
return didBlockScript;
}
void HTMLSourceTracker::start(const HTMLInputStream& input, HTMLTokenizer* tokenizer, HTMLToken& token)
{
if (token.type() == HTMLTokenTypes::Uninitialized) {
m_previousSource.clear();
if (tokenizer->numberOfBufferedCharacters())
m_previousSource = tokenizer->bufferedCharacters();
} else
m_previousSource.append(m_currentSource);
m_currentSource = input.current();
token.setBaseOffset(m_currentSource.numberOfCharactersConsumed() - m_previousSource.length());
}
void HTMLSourceTracker::start(SegmentedString& currentInput, HTMLTokenizer* tokenizer, HTMLToken& token)
{
if (token.type() == HTMLToken::Uninitialized) {
m_previousSource.clear();
if (tokenizer->numberOfBufferedCharacters())
m_previousSource = tokenizer->bufferedCharacters();
} else
m_previousSource.append(m_currentSource);
m_currentSource = currentInput;
token.setBaseOffset(m_currentSource.numberOfCharactersConsumed() - m_previousSource.length());
}
bool XSSAuditor::filterScriptToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, scriptTag));
if (eraseAttributeIfInjected(token, srcAttr, blankURL().string(), SrcLikeAttribute))
return true;
m_state = AfterScriptStartTag;
m_cachedSnippet = m_parser->sourceForToken(token);
return false;
}
bool XSSAuditor::filterObjectToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, objectTag));
bool didBlockScript = false;
if (isContainedInRequest(decodedSnippetForName(token))) {
didBlockScript |= eraseAttributeIfInjected(token, dataAttr, blankURL().string(), SrcLikeAttribute);
didBlockScript |= eraseAttributeIfInjected(token, typeAttr);
didBlockScript |= eraseAttributeIfInjected(token, classidAttr);
}
return didBlockScript;
}
bool XSSFilter::filterEmbedToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLToken::StartTag);
ASSERT(hasName(token, embedTag));
bool didBlockScript = false;
didBlockScript |= eraseAttributeIfInjected(token, srcAttr, blankURL().string());
didBlockScript |= eraseAttributeIfInjected(token, typeAttr);
return didBlockScript;
}
bool XSSAuditor::filterAppletToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, appletTag));
bool didBlockScript = false;
didBlockScript |= eraseAttributeIfInjected(token, codeAttr, String(), SrcLikeAttribute);
didBlockScript |= eraseAttributeIfInjected(token, objectAttr);
return didBlockScript;
}
bool XSSAuditor::filterScriptToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, scriptTag));
m_cachedDecodedSnippet = decodedSnippetForName(token);
m_shouldAllowCDATA = m_parser->tokenizer()->shouldAllowCDATA();
if (isContainedInRequest(decodedSnippetForName(token)))
return eraseAttributeIfInjected(token, srcAttr, blankURL().string(), SrcLikeAttribute);
return false;
}
bool XSSAuditor::filterTokenAfterScriptStartTag(HTMLToken& token)
{
ASSERT(m_state == AfterScriptStartTag);
m_state = Initial;
if (token.type() != HTMLTokenTypes::Character) {
ASSERT(token.type() == HTMLTokenTypes::EndTag || token.type() == HTMLTokenTypes::EndOfFile);
return false;
}
TextResourceDecoder* decoder = m_parser->document()->decoder();
if (isContainedInRequest(fullyDecodeString(m_cachedSnippet, decoder))) {
int start = 0;
int end = token.endIndex() - token.startIndex();
String snippet = snippetForJavaScript(snippetForRange(token, start, end));
if (isContainedInRequest(fullyDecodeString(snippet, decoder))) {
token.eraseCharacters();
token.appendToCharacter(' '); // Technically, character tokens can't be empty.
return true;
}
}
return false;
}
bool XSSAuditor::filterObjectToken(HTMLToken& token)
{
ASSERT(m_state == Initial);
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, objectTag));
bool didBlockScript = false;
didBlockScript |= eraseAttributeIfInjected(token, dataAttr, blankURL().string(), SrcLikeAttribute);
didBlockScript |= eraseAttributeIfInjected(token, typeAttr);
didBlockScript |= eraseAttributeIfInjected(token, classidAttr);
return didBlockScript;
}
// This is a regression test for crbug.com/619141
TEST(HTMLTokenizerTest, ZeroOffsetAttributeNameRange) {
HTMLParserOptions options;
std::unique_ptr<HTMLTokenizer> tokenizer = HTMLTokenizer::create(options);
HTMLToken token;
SegmentedString input("<script ");
EXPECT_FALSE(tokenizer->nextToken(input, token));
EXPECT_EQ(HTMLToken::StartTag, token.type());
SegmentedString input2("type='javascript'");
// Below should not fail ASSERT
EXPECT_FALSE(tokenizer->nextToken(input2, token));
}
void XSSAuditor::filterToken(HTMLToken& token)
{
if (m_state == Uninitialized)
init();
ASSERT(m_state == Initialized);
if (!m_isEnabled || m_xssProtection == XSSProtectionDisabled)
return;
bool didBlockScript = false;
if (token.type() == HTMLTokenTypes::StartTag)
didBlockScript = filterStartToken(token);
else if (m_scriptTagNestingLevel) {
if (token.type() == HTMLTokenTypes::Character)
didBlockScript = filterCharacterToken(token);
else if (token.type() == HTMLTokenTypes::EndTag)
filterEndToken(token);
}
if (didBlockScript) {
// FIXME: Consider using a more helpful console message.
DEFINE_STATIC_LOCAL(String, consoleMessage, ("Refused to execute a JavaScript script. Source code of script found within request.\n"));
m_parser->document()->addConsoleMessage(JSMessageSource, LogMessageType, ErrorMessageLevel, consoleMessage);
bool didBlockEntirePage = (m_xssProtection == XSSProtectionBlockEnabled);
if (didBlockEntirePage)
m_parser->document()->frame()->loader()->stopAllLoaders();
if (!m_notifiedClient) {
m_parser->document()->frame()->loader()->client()->didDetectXSS(m_parser->document()->url(), didBlockEntirePage);
m_notifiedClient = true;
}
if (didBlockEntirePage)
m_parser->document()->frame()->navigationScheduler()->scheduleLocationChange(m_parser->document()->securityOrigin(), blankURL(), String());
}
}
void HTMLDocumentParser::constructTreeFromHTMLToken(HTMLToken& rawToken)
{
AtomicHTMLToken token(rawToken);
// We clear the rawToken in case constructTreeFromAtomicToken
// synchronously re-enters the parser. We don't clear the token immedately
// for Character tokens because the AtomicHTMLToken avoids copying the
// characters by keeping a pointer to the underlying buffer in the
// HTMLToken. Fortunately, Character tokens can't cause us to re-enter
// the parser.
//
// FIXME: Stop clearing the rawToken once we start running the parser off
// the main thread or once we stop allowing synchronous JavaScript
// execution from parseAttribute.
if (rawToken.type() != HTMLToken::Character)
rawToken.clear();
m_treeBuilder->constructTree(&token);
if (!rawToken.isUninitialized()) {
ASSERT(rawToken.type() == HTMLToken::Character);
rawToken.clear();
}
}
bool XSSAuditor::filterParamToken(HTMLToken& token)
{
ASSERT(token.type() == HTMLTokenTypes::StartTag);
ASSERT(hasName(token, paramTag));
size_t indexOfNameAttribute;
if (!findAttributeWithName(token, nameAttr, indexOfNameAttribute))
return false;
const HTMLToken::Attribute& nameAttribute = token.attributes().at(indexOfNameAttribute);
String name = String(nameAttribute.m_value.data(), nameAttribute.m_value.size());
if (!HTMLParamElement::isURLParameter(name))
return false;
return eraseAttributeIfInjected(token, valueAttr, blankURL().string(), SrcLikeAttribute);
}
void TokenPreloadScanner::scan(const HTMLToken& token, Vector<std::unique_ptr<PreloadRequest>>& requests, Document& document)
{
switch (token.type()) {
case HTMLToken::Character:
if (!m_inStyle)
return;
m_cssScanner.scan(token.characters(), requests);
return;
case HTMLToken::EndTag: {
TagId tagId = tagIdFor(token.name());
#if ENABLE(TEMPLATE_ELEMENT)
if (tagId == TagId::Template) {
if (m_templateCount)
--m_templateCount;
return;
}
#endif
if (tagId == TagId::Style) {
if (m_inStyle)
m_cssScanner.reset();
m_inStyle = false;
} else if (tagId == TagId::Picture && !m_pictureSourceState.isEmpty())
m_pictureSourceState.removeLast();
return;
}
case HTMLToken::StartTag: {
#if ENABLE(TEMPLATE_ELEMENT)
if (m_templateCount)
return;
#endif
TagId tagId = tagIdFor(token.name());
#if ENABLE(TEMPLATE_ELEMENT)
if (tagId == TagId::Template) {
++m_templateCount;
return;
}
#endif
if (tagId == TagId::Style) {
m_inStyle = true;
return;
}
if (tagId == TagId::Base) {
// The first <base> element is the one that wins.
if (!m_predictedBaseElementURL.isEmpty())
return;
updatePredictedBaseURL(token);
return;
}
if (tagId == TagId::Picture) {
m_pictureSourceState.append(false);
return;
}
StartTagScanner scanner(tagId, m_deviceScaleFactor);
scanner.processAttributes(token.attributes(), document, m_pictureSourceState);
if (auto request = scanner.createPreloadRequest(m_predictedBaseElementURL))
requests.append(WTFMove(request));
return;
}
default:
return;
}
}
