From 46a8885ae486e238a39efa5119c2714f328b08e4 Mon Sep 17 00:00:00 2001 From: Mitch Curtis Date: Fri, 27 Sep 2013 12:32:28 +0200 Subject: [PATCH] Disallow deep or widely nested entity references. Nested references with a depth of 2 or greater will fail. References that partially expand to greater than 1024 characters will also fail. Change-Id: Id4e49d6f7cf51e3a247efdb4c6c7c9bd9b223f6e Reviewed-by: Richard J. Moore Reviewed-by: Lars Knoll From f1053d94f59f053ce4acad9320df14f1fbe4faac Mon Sep 17 00:00:00 2001 From: Mitch Curtis Date: Mon, 11 Nov 2013 14:27:40 +0100 Subject: [PATCH] Fully expand entities to ensure deep or widely nested ones fail parsing With 46a8885ae486e238a39efa5119c2714f328b08e4, we failed when parsing entities whose partially expanded size was greater than 1024 characters. That was not enough, so now we fully expand all entities. Amends 46a8885ae486e238a39efa5119c2714f328b08e4. Change-Id: Ie80720d7e04d825eb4eebf528140eb94806c02b1 Reviewed-by: Richard J. Moore Reviewed-by: Lars Knoll diff --git a/src/xml/sax/qxml.cpp b/src/xml/sax/qxml.cpp index 45c0f3e..e6d78d3 100644 --- a/src/xml/sax/qxml.cpp +++ b/src/xml/sax/qxml.cpp @@ -424,6 +424,10 @@ private: int stringValueLen; QString emptyStr; + // The limit to the amount of times the DTD parsing functions can be called + // for the DTD currently being parsed. + int dtdRecursionLimit; + const QString &string(); void stringClear(); void stringAddC(QChar); @@ -493,6 +497,8 @@ private: void parseFailed(ParseFunction where, int state); void pushParseState(ParseFunction function, int state); + bool isPartiallyExpandedEntityValueTooLarge(QString *errorMessage); + Q_DECLARE_PUBLIC(QXmlSimpleReader) QXmlSimpleReader *q_ptr; @@ -2757,6 +2763,8 @@ QXmlSimpleReaderPrivate::QXmlSimpleReaderPrivate(QXmlSimpleReader *reader) useNamespacePrefixes = false; reportWhitespaceCharData = true; reportEntities = false; + + dtdRecursionLimit = 2; } QXmlSimpleReaderPrivate::~QXmlSimpleReaderPrivate() @@ -5035,6 +5043,11 @@ bool QXmlSimpleReaderPrivate::parseDoctype() } break; case Mup: + if (dtdRecursionLimit > 0 && parameterEntities.size() > dtdRecursionLimit) { + reportParseError(QString::fromLatin1( + "DTD parsing exceeded recursion limit of %1.").arg(dtdRecursionLimit)); + return false; + } if (!parseMarkupdecl()) { parseFailed(&QXmlSimpleReaderPrivate::parseDoctype, state); return false; @@ -6644,6 +6657,37 @@ bool QXmlSimpleReaderPrivate::parseChoiceSeq() return false; } +bool QXmlSimpleReaderPrivate::isPartiallyExpandedEntityValueTooLarge(QString *errorMessage) +{ + const QString value = string(); + QMap referencedEntityCounts; + foreach (QString entityName, entities.keys()) { + for (int i = 0; i < value.size() && i != -1; ) { + i = value.indexOf(entityName, i); + if (i != -1) { + // The entityName we're currently trying to find + // was matched in this string; increase our count. + ++referencedEntityCounts[entityName]; + i += entityName.size(); + } + } + } + + foreach (QString entityName, referencedEntityCounts.keys()) { + const int timesReferenced = referencedEntityCounts[entityName]; + const QString entityValue = entities[entityName]; + if (entityValue.size() * timesReferenced > 1024) { + if (errorMessage) { + *errorMessage = QString::fromLatin1("The XML entity \"%1\"" + "expands too a string that is too large to process when " + "referencing \"%2\" %3 times.").arg(entityName).arg(entityName).arg(timesReferenced); + } + return true; + } + } + return false; +} + /* Parse a EntityDecl [70]. @@ -6738,6 +6782,15 @@ bool QXmlSimpleReaderPrivate::parseEntityDecl() switch (state) { case EValue: if ( !entityExist(name())) { + QString errorMessage; + if (isPartiallyExpandedEntityValueTooLarge(&errorMessage)) { + // The entity at entityName is entityValue.size() characters + // long in its unexpanded form, and was mentioned timesReferenced times, + // resulting in a string that would be greater than 1024 characters. + reportParseError(errorMessage); + return false; + } + entities.insert(name(), string()); if (declHnd) { if (!declHnd->internalEntityDecl(name(), string())) { diff --git a/src/xml/sax/qxml.cpp b/src/xml/sax/qxml.cpp index e6d78d3..f3a1e47 100644 --- a/src/xml/sax/qxml.cpp +++ b/src/xml/sax/qxml.cpp @@ -426,7 +426,9 @@ private: // The limit to the amount of times the DTD parsing functions can be called // for the DTD currently being parsed. - int dtdRecursionLimit; + static const int dtdRecursionLimit = 2; + // The maximum amount of characters an entity value may contain, after expansion. + static const int entityCharacterLimit = 1024; const QString &string(); void stringClear(); @@ -497,7 +499,7 @@ private: void parseFailed(ParseFunction where, int state); void pushParseState(ParseFunction function, int state); - bool isPartiallyExpandedEntityValueTooLarge(QString *errorMessage); + bool isExpandedEntityValueTooLarge(QString *errorMessage); Q_DECLARE_PUBLIC(QXmlSimpleReader) QXmlSimpleReader *q_ptr; @@ -2763,8 +2765,6 @@ QXmlSimpleReaderPrivate::QXmlSimpleReaderPrivate(QXmlSimpleReader *reader) useNamespacePrefixes = false; reportWhitespaceCharData = true; reportEntities = false; - - dtdRecursionLimit = 2; } QXmlSimpleReaderPrivate::~QXmlSimpleReaderPrivate() @@ -6657,30 +6657,43 @@ bool QXmlSimpleReaderPrivate::parseChoiceSeq() return false; } -bool QXmlSimpleReaderPrivate::isPartiallyExpandedEntityValueTooLarge(QString *errorMessage) +bool QXmlSimpleReaderPrivate::isExpandedEntityValueTooLarge(QString *errorMessage) { - const QString value = string(); - QMap referencedEntityCounts; - foreach (QString entityName, entities.keys()) { - for (int i = 0; i < value.size() && i != -1; ) { - i = value.indexOf(entityName, i); - if (i != -1) { - // The entityName we're currently trying to find - // was matched in this string; increase our count. - ++referencedEntityCounts[entityName]; - i += entityName.size(); + QMap literalEntitySizes; + // The entity at (QMap) times. + QMap > referencesToOtherEntities; + QMap expandedSizes; + + // For every entity, check how many times all entity names were referenced in its value. + foreach (QString toSearch, entities.keys()) { + // The amount of characters that weren't entity names, but literals, like 'X'. + QString leftOvers = entities.value(toSearch); + // How many times was entityName referenced by toSearch? + foreach (QString entityName, entities.keys()) { + for (int i = 0; i < leftOvers.size() && i != -1; ) { + i = leftOvers.indexOf(QString::fromLatin1("&%1;").arg(entityName), i); + if (i != -1) { + leftOvers.remove(i, entityName.size() + 2); + // The entityName we're currently trying to find was matched in this string; increase our count. + ++referencesToOtherEntities[toSearch][entityName]; + } } } + literalEntitySizes[toSearch] = leftOvers.size(); } - foreach (QString entityName, referencedEntityCounts.keys()) { - const int timesReferenced = referencedEntityCounts[entityName]; - const QString entityValue = entities[entityName]; - if (entityValue.size() * timesReferenced > 1024) { + foreach (QString entity, referencesToOtherEntities.keys()) { + expandedSizes[entity] = literalEntitySizes[entity]; + foreach (QString referenceTo, referencesToOtherEntities.value(entity).keys()) { + const int references = referencesToOtherEntities.value(entity).value(referenceTo); + // The total size of an entity's value is the expanded size of all of its referenced entities, plus its literal size. + expandedSizes[entity] += expandedSizes[referenceTo] * references + literalEntitySizes[referenceTo] * references; + } + + if (expandedSizes[entity] > entityCharacterLimit) { if (errorMessage) { - *errorMessage = QString::fromLatin1("The XML entity \"%1\"" - "expands too a string that is too large to process when " - "referencing \"%2\" %3 times.").arg(entityName).arg(entityName).arg(timesReferenced); + *errorMessage = QString::fromLatin1("The XML entity \"%1\" expands too a string that is too large to process (%2 characters > %3)."); + *errorMessage = (*errorMessage).arg(entity).arg(expandedSizes[entity]).arg(entityCharacterLimit); } return true; } @@ -6783,10 +6796,7 @@ bool QXmlSimpleReaderPrivate::parseEntityDecl() case EValue: if ( !entityExist(name())) { QString errorMessage; - if (isPartiallyExpandedEntityValueTooLarge(&errorMessage)) { - // The entity at entityName is entityValue.size() characters - // long in its unexpanded form, and was mentioned timesReferenced times, - // resulting in a string that would be greater than 1024 characters. + if (isExpandedEntityValueTooLarge(&errorMessage)) { reportParseError(errorMessage); return false; } -- 1.7