diff --git a/clang/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h b/clang/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h --- a/clang/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h +++ b/clang/include/clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h @@ -84,6 +84,8 @@ return Eng.getContext(); } + const ASTContext &getASTContext() const { return Eng.getContext(); } + const LangOptions &getLangOpts() const { return Eng.getContext().getLangOpts(); } diff --git a/clang/lib/StaticAnalyzer/Checkers/GenericTaintChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/GenericTaintChecker.cpp --- a/clang/lib/StaticAnalyzer/Checkers/GenericTaintChecker.cpp +++ b/clang/lib/StaticAnalyzer/Checkers/GenericTaintChecker.cpp @@ -22,15 +22,14 @@ #include "clang/StaticAnalyzer/Core/BugReporter/BugType.h" #include "clang/StaticAnalyzer/Core/Checker.h" #include "clang/StaticAnalyzer/Core/CheckerManager.h" +#include "clang/StaticAnalyzer/Core/PathSensitive/CallDescription.h" #include "clang/StaticAnalyzer/Core/PathSensitive/CallEvent.h" #include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h" #include "clang/StaticAnalyzer/Core/PathSensitive/ProgramStateTrait.h" #include "llvm/Support/YAMLTraits.h" -#include #include #include -#include #include using namespace clang; @@ -38,577 +37,651 @@ using namespace taint; namespace { -class GenericTaintChecker : public Checker { -public: - static void *getTag() { - static int Tag; - return &Tag; + +class GenericTaintChecker; + +/// Check for CWE-134: Uncontrolled Format String. +constexpr llvm::StringLiteral MsgUncontrolledFormatString = + "Untrusted data is used as a format string " + "(CWE-134: Uncontrolled Format String)"; + +/// Check for: +/// CERT/STR02-C. "Sanitize data passed to complex subsystems" +/// CWE-78, "Failure to Sanitize Data into an OS Command" +constexpr llvm::StringLiteral MsgSanitizeSystemArgs = + "Untrusted data is passed to a system call " + "(CERT/STR02-C. Sanitize data passed to complex subsystems)"; + +/// Check if tainted data is used as a buffer size in strn.. functions, +/// and allocators. +constexpr llvm::StringLiteral MsgTaintedBufferSize = + "Untrusted data is used to specify the buffer size " + "(CERT/STR31-C. Guarantee that storage for strings has sufficient space " + "for character data and the null terminator)"; + +/// Check if tainted data is used as a custom sink's parameter. +constexpr llvm::StringLiteral MsgCustomSink = + "Untrusted data is passed to a user-defined sink"; + +using ArgIdxTy = int; +using ArgVecTy = llvm::SmallVector; + +/// Denotes the return value. +constexpr ArgIdxTy ReturnValueIndex{-1}; + +static ArgIdxTy fromArgumentCount(unsigned Count) { + assert(Count <= + static_cast(std::numeric_limits::max()) && + "ArgIdxTy is not large enough to represent the number of arguments."); + return Count; +} + +/// Check if the region the expression evaluates to is the standard input, +/// and thus, is tainted. +/// FIXME: Move this to Taint.cpp. +bool isStdin(SVal Val, const ASTContext &ACtx) { + // FIXME: What if Val is NonParamVarRegion? + + // The region should be symbolic, we do not know it's value. + const auto *SymReg = dyn_cast_or_null(Val.getAsRegion()); + if (!SymReg) + return false; + + // Get it's symbol and find the declaration region it's pointing to. + const auto *Sm = dyn_cast(SymReg->getSymbol()); + if (!Sm) + return false; + const auto *DeclReg = dyn_cast(Sm->getRegion()); + if (!DeclReg) + return false; + + // This region corresponds to a declaration, find out if it's a global/extern + // variable named stdin with the proper type. + if (const auto *D = dyn_cast_or_null(DeclReg->getDecl())) { + D = D->getCanonicalDecl(); + // FIXME: This should look for an exact match. + if (D->getName().contains("stdin") && D->isExternC()) { + const QualType FILETy = ACtx.getFILEType().getCanonicalType(); + const QualType Ty = D->getType().getCanonicalType(); + + if (Ty->isPointerType()) + return Ty->getPointeeType() == FILETy; + } } + return false; +} - void checkPreCall(const CallEvent &Call, CheckerContext &C) const; - void checkPostCall(const CallEvent &Call, CheckerContext &C) const; +SVal getPointeeOf(const CheckerContext &C, Loc LValue) { + const QualType ArgTy = LValue.getType(C.getASTContext()); + if (!ArgTy->isPointerType() || !ArgTy->getPointeeType()->isVoidType()) + return C.getState()->getSVal(LValue); - void printState(raw_ostream &Out, ProgramStateRef State, const char *NL, - const char *Sep) const override; + // Do not dereference void pointers. Treat them as byte pointers instead. + // FIXME: we might want to consider more than just the first byte. + return C.getState()->getSVal(LValue, C.getASTContext().CharTy); +} + +/// Given a pointer/reference argument, return the value it refers to. +Optional getPointeeOf(const CheckerContext &C, SVal Arg) { + if (auto LValue = Arg.getAs()) + return getPointeeOf(C, *LValue); + return None; +} - using ArgVector = SmallVector; - using SignedArgVector = SmallVector; +/// Given a pointer, return the SVal of its pointee or if it is tainted, +/// otherwise return the pointer's SVal if tainted. +/// Also considers stdin as a taint source. +Optional getTaintedPointeeOrPointer(const CheckerContext &C, SVal Arg) { + const ProgramStateRef State = C.getState(); - enum class VariadicType { None, Src, Dst }; + if (auto Pointee = getPointeeOf(C, Arg)) + if (isTainted(State, *Pointee)) // FIXME: isTainted(...) ? Pointee : None; + return Pointee; - /// Used to parse the configuration file. - struct TaintConfiguration { - using NameScopeArgs = std::tuple; - - struct Propagation { - std::string Name; - std::string Scope; - ArgVector SrcArgs; - SignedArgVector DstArgs; - VariadicType VarType; - unsigned VarIndex; - }; - - std::vector Propagations; - std::vector Filters; - std::vector Sinks; - - TaintConfiguration() = default; - TaintConfiguration(const TaintConfiguration &) = default; - TaintConfiguration(TaintConfiguration &&) = default; - TaintConfiguration &operator=(const TaintConfiguration &) = default; - TaintConfiguration &operator=(TaintConfiguration &&) = default; - }; + if (isTainted(State, Arg)) + return Arg; + + // FIXME: This should be done by the isTainted() API. + if (isStdin(Arg, C.getASTContext())) + return Arg; + + return None; +} - /// Convert SignedArgVector to ArgVector. - ArgVector convertToArgVector(CheckerManager &Mgr, const std::string &Option, - const SignedArgVector &Args); +bool isTaintedOrPointsToTainted(const Expr *E, const ProgramStateRef &State, + CheckerContext &C) { + return getTaintedPointeeOrPointer(C, C.getSVal(E)).hasValue(); +} - /// Parse the config. - void parseConfiguration(CheckerManager &Mgr, const std::string &Option, - TaintConfiguration &&Config); +/// ArgSet is used to describe arguments relevant for taint detection or +/// taint application. A discrete set of argument indexes and a variadic +/// argument list signified by a starting index are supported. +class ArgSet { +public: + ArgSet() = default; + ArgSet(ArgVecTy &&DiscreteArgs, Optional VariadicIndex = None) + : DiscreteArgs(std::move(DiscreteArgs)), + VariadicIndex(std::move(VariadicIndex)) {} - static const unsigned InvalidArgIndex{std::numeric_limits::max()}; - /// Denotes the return vale. - static const unsigned ReturnValueIndex{std::numeric_limits::max() - - 1}; + bool contains(ArgIdxTy ArgIdx) const { + if (llvm::is_contained(DiscreteArgs, ArgIdx)) + return true; -private: - mutable std::unique_ptr BT; - void initBugType() const { - if (!BT) - BT = std::make_unique(this, "Use of Untrusted Data", - "Untrusted Data"); + return VariadicIndex && ArgIdx >= *VariadicIndex; } - struct FunctionData { - FunctionData() = delete; - FunctionData(const FunctionDecl *FDecl, StringRef Name, - std::string FullName) - : FDecl(FDecl), Name(Name), FullName(std::move(FullName)) {} - FunctionData(const FunctionData &) = default; - FunctionData(FunctionData &&) = default; - FunctionData &operator=(const FunctionData &) = delete; - FunctionData &operator=(FunctionData &&) = delete; - - static Optional create(const CallEvent &Call, - const CheckerContext &C) { - if (!Call.getDecl()) - return None; - - const FunctionDecl *FDecl = Call.getDecl()->getAsFunction(); - if (!FDecl || (FDecl->getKind() != Decl::Function && - FDecl->getKind() != Decl::CXXMethod)) - return None; - - StringRef Name = C.getCalleeName(FDecl); - std::string FullName = FDecl->getQualifiedNameAsString(); - if (Name.empty() || FullName.empty()) - return None; - - return FunctionData{FDecl, Name, std::move(FullName)}; - } + bool isEmpty() const { return DiscreteArgs.empty() && !VariadicIndex; } - bool isInScope(StringRef Scope) const { - return StringRef(FullName).startswith(Scope); + ArgVecTy ArgsUpTo(ArgIdxTy LastArgIdx) const { + ArgVecTy Args; + for (ArgIdxTy I = ReturnValueIndex; I <= LastArgIdx; ++I) { + if (contains(I)) + Args.push_back(I); } + return Args; + } - const FunctionDecl *const FDecl; - const StringRef Name; - const std::string FullName; - }; +private: + ArgVecTy DiscreteArgs; + Optional VariadicIndex; +}; - /// Catch taint related bugs. Check if tainted data is passed to a - /// system call etc. Returns true on matching. - bool checkPre(const CallEvent &Call, const FunctionData &FData, - CheckerContext &C) const; +/// A struct used to specify taint propagation rules for a function. +/// +/// If any of the possible taint source arguments is tainted, all of the +/// destination arguments should also be tainted. If ReturnValueIndex is added +/// to the dst list, the return value will be tainted. +class GenericTaintRule { + /// Arguments which are taints sinks and should be checked, and a report + /// should be emitted if taint reaches these. + ArgSet SinkArgs; + /// Arguments which should be sanitized on function return. + ArgSet FilterArgs; + /// Arguments which can participate in taint propagationa. If any of the + /// arguments in PropSrcArgs is tainted, all arguments in PropDstArgs should + /// be tainted. + ArgSet PropSrcArgs; + ArgSet PropDstArgs; + + /// A message that explains why the call is sensitive to taint. + Optional SinkMsg; + + GenericTaintRule() = default; + + GenericTaintRule(ArgSet &&Sink, ArgSet &&Filter, ArgSet &&Src, ArgSet &&Dst, + Optional SinkMsg = None) + : SinkArgs(std::move(Sink)), FilterArgs(std::move(Filter)), + PropSrcArgs(std::move(Src)), PropDstArgs(std::move(Dst)), + SinkMsg(SinkMsg) {} - /// Add taint sources on a pre-visit. Returns true on matching. - bool addSourcesPre(const CallEvent &Call, const FunctionData &FData, - CheckerContext &C) const; +public: + /// Make a rule that reports a warning if taint reaches any of \p FilterArgs + /// arguments. + static GenericTaintRule Sink(ArgSet &&SinkArgs, + Optional Msg = None) { + return {std::move(SinkArgs), {}, {}, {}, Msg}; + } - /// Mark filter's arguments not tainted on a pre-visit. Returns true on - /// matching. - bool addFiltersPre(const CallEvent &Call, const FunctionData &FData, - CheckerContext &C) const; + /// Make a rule that sanitizes all FilterArgs arguments. + static GenericTaintRule Filter(ArgSet &&FilterArgs) { + return {{}, std::move(FilterArgs), {}, {}}; + } - /// Propagate taint generated at pre-visit. Returns true on matching. - static bool propagateFromPre(const CallEvent &Call, CheckerContext &C); + /// Make a rule that unconditionally taints all Args. + /// If Func is provided, it must also return true for taint to propagate. + static GenericTaintRule Source(ArgSet &&SourceArgs) { + return {{}, {}, {}, std::move(SourceArgs)}; + } - /// Check if the region the expression evaluates to is the standard input, - /// and thus, is tainted. - static bool isStdin(const Expr *E, CheckerContext &C); + /// Make a rule that taints all PropDstArgs if any of PropSrcArgs is tainted. + static GenericTaintRule Prop(ArgSet &&SrcArgs, ArgSet &&DstArgs) { + return {{}, {}, std::move(SrcArgs), std::move(DstArgs)}; + } - /// Given a pointer argument, return the value it points to. - static Optional getPointeeOf(CheckerContext &C, const Expr *Arg); + /// Make a rule that taints all PropDstArgs if any of PropSrcArgs is tainted. + static GenericTaintRule SinkProp(ArgSet &&SinkArgs, ArgSet &&SrcArgs, + ArgSet &&DstArgs, + Optional Msg = None) { + return { + std::move(SinkArgs), {}, std::move(SrcArgs), std::move(DstArgs), Msg}; + } - /// Check for CWE-134: Uncontrolled Format String. - static constexpr llvm::StringLiteral MsgUncontrolledFormatString = - "Untrusted data is used as a format string " - "(CWE-134: Uncontrolled Format String)"; - bool checkUncontrolledFormatString(const CallEvent &Call, - CheckerContext &C) const; + /// Process a function which could either be a taint source, a taint sink, a + /// taint filter or a taint propagator. + void process(const GenericTaintChecker &Checker, const CallEvent &Call, + CheckerContext &C) const; - /// Check for: - /// CERT/STR02-C. "Sanitize data passed to complex subsystems" - /// CWE-78, "Failure to Sanitize Data into an OS Command" - static constexpr llvm::StringLiteral MsgSanitizeSystemArgs = - "Untrusted data is passed to a system call " - "(CERT/STR02-C. Sanitize data passed to complex subsystems)"; - bool checkSystemCall(const CallEvent &Call, StringRef Name, - CheckerContext &C) const; - - /// Check if tainted data is used as a buffer size ins strn.. functions, - /// and allocators. - static constexpr llvm::StringLiteral MsgTaintedBufferSize = - "Untrusted data is used to specify the buffer size " - "(CERT/STR31-C. Guarantee that storage for strings has sufficient space " - "for character data and the null terminator)"; - bool checkTaintedBufferSize(const CallEvent &Call, CheckerContext &C) const; - - /// Check if tainted data is used as a custom sink's parameter. - static constexpr llvm::StringLiteral MsgCustomSink = - "Untrusted data is passed to a user-defined sink"; - bool checkCustomSinks(const CallEvent &Call, const FunctionData &FData, - CheckerContext &C) const; + /// Handles the resolution of indexes of type ArgIdxTy to Expr*-s. + static const Expr *GetArgExpr(ArgIdxTy ArgIdx, const CallEvent &Call) { + return ArgIdx == ReturnValueIndex ? Call.getOriginExpr() + : Call.getArgExpr(ArgIdx); + }; - /// Generate a report if the expression is tainted or points to tainted data. - bool generateReportIfTainted(const Expr *E, StringRef Msg, - CheckerContext &C) const; + /// Functions for custom taintedness propagation. + static bool UntrustedEnv(CheckerContext &C); +}; + +using RuleLookupTy = CallDescriptionMap; + +/// Used to parse the configuration file. +struct TaintConfiguration { + using NameScopeArgs = std::tuple; + enum class VariadicType { None, Src, Dst }; + + struct Common { + std::string Name; + std::string Scope; + }; + + struct Sink : Common { + ArgVecTy SinkArgs; + }; + + struct Filter : Common { + ArgVecTy FilterArgs; + }; - struct TaintPropagationRule; - template - using ConfigDataMap = - std::unordered_multimap>; - using NameRuleMap = ConfigDataMap; - using NameArgMap = ConfigDataMap; - - /// Find a function with the given name and scope. Returns the first match - /// or the end of the map. - template - static auto findFunctionInConfig(const ConfigDataMap &Map, - const FunctionData &FData); - - /// A struct used to specify taint propagation rules for a function. - /// - /// If any of the possible taint source arguments is tainted, all of the - /// destination arguments should also be tainted. Use InvalidArgIndex in the - /// src list to specify that all of the arguments can introduce taint. Use - /// InvalidArgIndex in the dst arguments to signify that all the non-const - /// pointer and reference arguments might be tainted on return. If - /// ReturnValueIndex is added to the dst list, the return value will be - /// tainted. - struct TaintPropagationRule { - using PropagationFuncType = bool (*)(bool IsTainted, const CallEvent &Call, - CheckerContext &C); - - /// List of arguments which can be taint sources and should be checked. - ArgVector SrcArgs; - /// List of arguments which should be tainted on function return. - ArgVector DstArgs; - /// Index for the first variadic parameter if exist. - unsigned VariadicIndex; - /// Show when a function has variadic parameters. If it has, it marks all - /// of them as source or destination. + struct Propagation : Common { + ArgVecTy SrcArgs; + ArgVecTy DstArgs; VariadicType VarType; - /// Special function for tainted source determination. If defined, it can - /// override the default behavior. - PropagationFuncType PropagationFunc; - - TaintPropagationRule() - : VariadicIndex(InvalidArgIndex), VarType(VariadicType::None), - PropagationFunc(nullptr) {} - - TaintPropagationRule(ArgVector &&Src, ArgVector &&Dst, - VariadicType Var = VariadicType::None, - unsigned VarIndex = InvalidArgIndex, - PropagationFuncType Func = nullptr) - : SrcArgs(std::move(Src)), DstArgs(std::move(Dst)), - VariadicIndex(VarIndex), VarType(Var), PropagationFunc(Func) {} - - /// Get the propagation rule for a given function. - static TaintPropagationRule - getTaintPropagationRule(const NameRuleMap &CustomPropagations, - const FunctionData &FData, CheckerContext &C); - - void addSrcArg(unsigned A) { SrcArgs.push_back(A); } - void addDstArg(unsigned A) { DstArgs.push_back(A); } - - bool isNull() const { - return SrcArgs.empty() && DstArgs.empty() && - VariadicType::None == VarType; - } + ArgIdxTy VarIndex; + }; - bool isDestinationArgument(unsigned ArgNum) const { - return llvm::is_contained(DstArgs, ArgNum); - } + std::vector Propagations; + std::vector Filters; + std::vector Sinks; - static bool isTaintedOrPointsToTainted(const Expr *E, - const ProgramStateRef &State, - CheckerContext &C) { - if (isTainted(State, E, C.getLocationContext()) || isStdin(E, C)) - return true; + TaintConfiguration() = default; + TaintConfiguration(const TaintConfiguration &) = default; + TaintConfiguration(TaintConfiguration &&) = default; + TaintConfiguration &operator=(const TaintConfiguration &) = default; + TaintConfiguration &operator=(TaintConfiguration &&) = default; +}; - if (!E->getType().getTypePtr()->isPointerType()) - return false; +struct GenericTaintRuleParser { + GenericTaintRuleParser(CheckerManager &Mgr) : Mgr(Mgr) {} + /// Container type used to gather call identification objects grouped into + /// pairs with their corresponding taint rules. It is temporary as it is used + /// to finally initialize RuleLookupTy, which is considered to be immutable. + using RulesContTy = std::vector>; + RulesContTy parseConfiguration(const std::string &Option, + TaintConfiguration &&Config) const; - Optional V = getPointeeOf(C, E); - return (V && isTainted(State, *V)); - } +private: + using NamePartsTy = llvm::SmallVector, 2>; - /// Pre-process a function which propagates taint according to the - /// taint rule. - ProgramStateRef process(const CallEvent &Call, CheckerContext &C) const; + /// Validate part of the configuration, which contains a list of argument + /// indexes. + void validateArgVector(const std::string &Option, const ArgVecTy &Args) const; - // Functions for custom taintedness propagation. - static bool postSocket(bool IsTainted, const CallEvent &Call, - CheckerContext &C); - }; + template static NamePartsTy parseNameParts(const Config &C); - /// Defines a map between the propagation function's name, scope - /// and TaintPropagationRule. - NameRuleMap CustomPropagations; + // Takes the config and creates a CallDescription for it and associates a Rule + // with that. + template + static void consumeRulesFromConfig(const Config &C, GenericTaintRule &&Rule, + RulesContTy &Rules); - /// Defines a map between the filter function's name, scope and filtering - /// args. - NameArgMap CustomFilters; + void parseConfig(const std::string &Option, TaintConfiguration::Sink &&P, + RulesContTy &Rules) const; + void parseConfig(const std::string &Option, TaintConfiguration::Filter &&P, + RulesContTy &Rules) const; + void parseConfig(const std::string &Option, + TaintConfiguration::Propagation &&P, + RulesContTy &Rules) const; - /// Defines a map between the sink function's name, scope and sinking args. - NameArgMap CustomSinks; + CheckerManager &Mgr; }; -const unsigned GenericTaintChecker::ReturnValueIndex; -const unsigned GenericTaintChecker::InvalidArgIndex; +class GenericTaintChecker : public Checker { +public: + static void *getTag() { + static int Tag; + return &Tag; + } -// FIXME: these lines can be removed in C++17 -constexpr llvm::StringLiteral GenericTaintChecker::MsgUncontrolledFormatString; -constexpr llvm::StringLiteral GenericTaintChecker::MsgSanitizeSystemArgs; -constexpr llvm::StringLiteral GenericTaintChecker::MsgTaintedBufferSize; -constexpr llvm::StringLiteral GenericTaintChecker::MsgCustomSink; -} // end of anonymous namespace + void checkPreCall(const CallEvent &Call, CheckerContext &C) const; + void checkPostCall(const CallEvent &Call, CheckerContext &C) const; -using TaintConfig = GenericTaintChecker::TaintConfiguration; + void printState(raw_ostream &Out, ProgramStateRef State, const char *NL, + const char *Sep) const override; -LLVM_YAML_IS_SEQUENCE_VECTOR(TaintConfig::Propagation) -LLVM_YAML_IS_SEQUENCE_VECTOR(TaintConfig::NameScopeArgs) + /// Generate a report if the expression is tainted or points to tainted data. + bool generateReportIfTainted(const Expr *E, StringRef Msg, + CheckerContext &C) const; + +private: + const BugType BT{this, "Use of Untrusted Data", "Untrusted Data"}; + + bool checkUncontrolledFormatString(const CallEvent &Call, + CheckerContext &C) const; + + void taintUnsafeSocketProtocol(const CallEvent &Call, + CheckerContext &C) const; + + /// Default taint rules are initilized with the help of a CheckerContext to + /// access the names of built-in functions like memcpy. + void initTaintRules(CheckerContext &C) const; + + /// CallDescription currently cannot restrict matches to the global namespace + /// only, which is why multiple CallDescriptionMaps are used, as we want to + /// disambiguate global C functions from functions inside user-defined + /// namespaces. + // TODO: Remove separation to simplify matching logic once CallDescriptions + // are more expressive. + + mutable Optional StaticTaintRules; + mutable Optional DynamicTaintRules; +}; +} // end of anonymous namespace + +/// YAML serialization mapping. +LLVM_YAML_IS_SEQUENCE_VECTOR(TaintConfiguration::Sink) +LLVM_YAML_IS_SEQUENCE_VECTOR(TaintConfiguration::Filter) +LLVM_YAML_IS_SEQUENCE_VECTOR(TaintConfiguration::Propagation) namespace llvm { namespace yaml { -template <> struct MappingTraits { - static void mapping(IO &IO, TaintConfig &Config) { +template <> struct MappingTraits { + static void mapping(IO &IO, TaintConfiguration &Config) { IO.mapOptional("Propagations", Config.Propagations); IO.mapOptional("Filters", Config.Filters); IO.mapOptional("Sinks", Config.Sinks); } }; -template <> struct MappingTraits { - static void mapping(IO &IO, TaintConfig::Propagation &Propagation) { +template <> struct MappingTraits { + static void mapping(IO &IO, TaintConfiguration::Sink &Sink) { + IO.mapRequired("Name", Sink.Name); + IO.mapOptional("Scope", Sink.Scope); + IO.mapRequired("Args", Sink.SinkArgs); + } +}; + +template <> struct MappingTraits { + static void mapping(IO &IO, TaintConfiguration::Filter &Filter) { + IO.mapRequired("Name", Filter.Name); + IO.mapOptional("Scope", Filter.Scope); + IO.mapRequired("Args", Filter.FilterArgs); + } +}; + +template <> struct MappingTraits { + static void mapping(IO &IO, TaintConfiguration::Propagation &Propagation) { IO.mapRequired("Name", Propagation.Name); IO.mapOptional("Scope", Propagation.Scope); IO.mapOptional("SrcArgs", Propagation.SrcArgs); IO.mapOptional("DstArgs", Propagation.DstArgs); - IO.mapOptional("VariadicType", Propagation.VarType, - GenericTaintChecker::VariadicType::None); - IO.mapOptional("VariadicIndex", Propagation.VarIndex, - GenericTaintChecker::InvalidArgIndex); - } -}; - -template <> struct ScalarEnumerationTraits { - static void enumeration(IO &IO, GenericTaintChecker::VariadicType &Value) { - IO.enumCase(Value, "None", GenericTaintChecker::VariadicType::None); - IO.enumCase(Value, "Src", GenericTaintChecker::VariadicType::Src); - IO.enumCase(Value, "Dst", GenericTaintChecker::VariadicType::Dst); + IO.mapOptional("VariadicType", Propagation.VarType); + IO.mapOptional("VariadicIndex", Propagation.VarIndex); } }; -template <> struct MappingTraits { - static void mapping(IO &IO, TaintConfig::NameScopeArgs &NSA) { - IO.mapRequired("Name", std::get<0>(NSA)); - IO.mapOptional("Scope", std::get<1>(NSA)); - IO.mapRequired("Args", std::get<2>(NSA)); +template <> struct ScalarEnumerationTraits { + static void enumeration(IO &IO, TaintConfiguration::VariadicType &Value) { + IO.enumCase(Value, "None", TaintConfiguration::VariadicType::None); + IO.enumCase(Value, "Src", TaintConfiguration::VariadicType::Src); + IO.enumCase(Value, "Dst", TaintConfiguration::VariadicType::Dst); } }; } // namespace yaml } // namespace llvm /// A set which is used to pass information from call pre-visit instruction -/// to the call post-visit. The values are unsigned integers, which are either +/// to the call post-visit. The values are signed integers, which are either /// ReturnValueIndex, or indexes of the pointer/reference argument, which /// points to data, which should be tainted on return. -REGISTER_SET_WITH_PROGRAMSTATE(TaintArgsOnPostVisit, unsigned) - -GenericTaintChecker::ArgVector -GenericTaintChecker::convertToArgVector(CheckerManager &Mgr, - const std::string &Option, - const SignedArgVector &Args) { - ArgVector Result; - for (int Arg : Args) { - if (Arg == -1) - Result.push_back(ReturnValueIndex); - else if (Arg < -1) { - Result.push_back(InvalidArgIndex); +REGISTER_SET_WITH_PROGRAMSTATE(TaintArgsOnPostVisit, ArgIdxTy) + +void GenericTaintRuleParser::validateArgVector(const std::string &Option, + const ArgVecTy &Args) const { + for (ArgIdxTy Arg : Args) { + if (Arg < ReturnValueIndex) { Mgr.reportInvalidCheckerOptionValue( - this, Option, + Mgr.getChecker(), Option, "an argument number for propagation rules greater or equal to -1"); - } else - Result.push_back(static_cast(Arg)); + } } - return Result; } -void GenericTaintChecker::parseConfiguration(CheckerManager &Mgr, - const std::string &Option, - TaintConfiguration &&Config) { - for (auto &P : Config.Propagations) { - GenericTaintChecker::CustomPropagations.emplace( - P.Name, - std::make_pair(P.Scope, TaintPropagationRule{ - std::move(P.SrcArgs), - convertToArgVector(Mgr, Option, P.DstArgs), - P.VarType, P.VarIndex})); +template +GenericTaintRuleParser::NamePartsTy +GenericTaintRuleParser::parseNameParts(const Config &C) { + NamePartsTy NameParts; + if (!C.Scope.empty()) { + // If the Scope argument contains multiple "::" parts, those are considered + // namespace identifiers. + llvm::SmallVector NSParts; + StringRef{C.Scope}.split(NSParts, "::", /*MaxSplit*/ -1, + /*KeepEmpty*/ false); + NameParts.append(NSParts.begin(), NSParts.end()); } + NameParts.emplace_back(C.Name); + return NameParts; +} - for (auto &F : Config.Filters) { - GenericTaintChecker::CustomFilters.emplace( - std::get<0>(F), - std::make_pair(std::move(std::get<1>(F)), std::move(std::get<2>(F)))); - } +template +void GenericTaintRuleParser::consumeRulesFromConfig(const Config &C, + GenericTaintRule &&Rule, + RulesContTy &Rules) { + NamePartsTy NameParts = parseNameParts(C); + llvm::SmallVector CallDescParts{NameParts.size()}; + llvm::transform(NameParts, CallDescParts.begin(), + [](SmallString<32> &S) { return S.c_str(); }); + Rules.template emplace_back(CallDescParts, std::move(Rule)); +} - for (auto &S : Config.Sinks) { - GenericTaintChecker::CustomSinks.emplace( - std::get<0>(S), - std::make_pair(std::move(std::get<1>(S)), std::move(std::get<2>(S)))); - } +void GenericTaintRuleParser::parseConfig(const std::string &Option, + TaintConfiguration::Sink &&S, + RulesContTy &Rules) const { + validateArgVector(Option, S.SinkArgs); + consumeRulesFromConfig(S, GenericTaintRule::Sink(std::move(S.SinkArgs)), + Rules); } -template -auto GenericTaintChecker::findFunctionInConfig(const ConfigDataMap &Map, - const FunctionData &FData) { - auto Range = Map.equal_range(std::string(FData.Name)); - auto It = - std::find_if(Range.first, Range.second, [&FData](const auto &Entry) { - const auto &Value = Entry.second; - StringRef Scope = Value.first; - return Scope.empty() || FData.isInScope(Scope); - }); - return It != Range.second ? It : Map.end(); +void GenericTaintRuleParser::parseConfig(const std::string &Option, + TaintConfiguration::Filter &&S, + RulesContTy &Rules) const { + validateArgVector(Option, S.FilterArgs); + consumeRulesFromConfig(S, GenericTaintRule::Filter(std::move(S.FilterArgs)), + Rules); } -GenericTaintChecker::TaintPropagationRule -GenericTaintChecker::TaintPropagationRule::getTaintPropagationRule( - const NameRuleMap &CustomPropagations, const FunctionData &FData, - CheckerContext &C) { - // TODO: Currently, we might lose precision here: we always mark a return - // value as tainted even if it's just a pointer, pointing to tainted data. +void GenericTaintRuleParser::parseConfig(const std::string &Option, + TaintConfiguration::Propagation &&P, + RulesContTy &Rules) const { + validateArgVector(Option, P.SrcArgs); + validateArgVector(Option, P.DstArgs); + bool IsSrcVariadic = P.VarType == TaintConfiguration::VariadicType::Src; + bool IsDstVariadic = P.VarType == TaintConfiguration::VariadicType::Dst; + Optional JustVarIndex = P.VarIndex; + + ArgSet SrcDesc(std::move(P.SrcArgs), IsSrcVariadic ? JustVarIndex : None); + ArgSet DstDesc(std::move(P.DstArgs), IsDstVariadic ? JustVarIndex : None); + + consumeRulesFromConfig( + P, GenericTaintRule::Prop(std::move(SrcDesc), std::move(DstDesc)), Rules); +} + +GenericTaintRuleParser::RulesContTy +GenericTaintRuleParser::parseConfiguration(const std::string &Option, + TaintConfiguration &&Config) const { + + RulesContTy Rules; + + for (auto &F : Config.Filters) + parseConfig(Option, std::move(F), Rules); + + for (auto &S : Config.Sinks) + parseConfig(Option, std::move(S), Rules); + for (auto &P : Config.Propagations) + parseConfig(Option, std::move(P), Rules); + + return Rules; +} + +void GenericTaintChecker::initTaintRules(CheckerContext &C) const { // Check for exact name match for functions without builtin substitutes. // Use qualified name, because these are C functions without namespace. - TaintPropagationRule Rule = - llvm::StringSwitch(FData.FullName) - // Source functions - // TODO: Add support for vfscanf & family. - .Case("fdopen", {{}, {ReturnValueIndex}}) - .Case("fopen", {{}, {ReturnValueIndex}}) - .Case("freopen", {{}, {ReturnValueIndex}}) - .Case("getch", {{}, {ReturnValueIndex}}) - .Case("getchar", {{}, {ReturnValueIndex}}) - .Case("getchar_unlocked", {{}, {ReturnValueIndex}}) - .Case("gets", {{}, {0, ReturnValueIndex}}) - .Case("scanf", {{}, {}, VariadicType::Dst, 1}) - .Case("socket", {{}, - {ReturnValueIndex}, - VariadicType::None, - InvalidArgIndex, - &TaintPropagationRule::postSocket}) - .Case("wgetch", {{}, {ReturnValueIndex}}) - // Propagating functions - .Case("atoi", {{0}, {ReturnValueIndex}}) - .Case("atol", {{0}, {ReturnValueIndex}}) - .Case("atoll", {{0}, {ReturnValueIndex}}) - .Case("fgetc", {{0}, {ReturnValueIndex}}) - .Case("fgetln", {{0}, {ReturnValueIndex}}) - .Case("fgets", {{2}, {0, ReturnValueIndex}}) - .Case("fscanf", {{0}, {}, VariadicType::Dst, 2}) - .Case("sscanf", {{0}, {}, VariadicType::Dst, 2}) - .Case("getc", {{0}, {ReturnValueIndex}}) - .Case("getc_unlocked", {{0}, {ReturnValueIndex}}) - .Case("getdelim", {{3}, {0}}) - .Case("getline", {{2}, {0}}) - .Case("getw", {{0}, {ReturnValueIndex}}) - .Case("pread", {{0, 1, 2, 3}, {1, ReturnValueIndex}}) - .Case("read", {{0, 2}, {1, ReturnValueIndex}}) - .Case("strchr", {{0}, {ReturnValueIndex}}) - .Case("strrchr", {{0}, {ReturnValueIndex}}) - .Case("tolower", {{0}, {ReturnValueIndex}}) - .Case("toupper", {{0}, {ReturnValueIndex}}) - .Default({}); - - if (!Rule.isNull()) - return Rule; + + if (StaticTaintRules || DynamicTaintRules) + return; + + using RulesConstructionTy = + std::vector>; + using TR = GenericTaintRule; + + const Builtin::Context &BI = C.getASTContext().BuiltinInfo; + + RulesConstructionTy GlobalCRules{ + // Sources + {{"fdopen"}, TR::Source({{ReturnValueIndex}})}, + {{"fopen"}, TR::Source({{ReturnValueIndex}})}, + {{"freopen"}, TR::Source({{ReturnValueIndex}})}, + {{"getch"}, TR::Source({{ReturnValueIndex}})}, + {{"getchar"}, TR::Source({{ReturnValueIndex}})}, + {{"getchar_unlocked"}, TR::Source({{ReturnValueIndex}})}, + {{"gets"}, TR::Source({{0}, ReturnValueIndex})}, + {{"scanf"}, TR::Source({{}, 1})}, + {{"wgetch"}, TR::Source({{}, ReturnValueIndex})}, + + // Props + {{"atoi"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"atol"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"atoll"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"fgetc"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"fgetln"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"fgets"}, TR::Prop({{2}}, {{0}, ReturnValueIndex})}, + {{"fscanf"}, TR::Prop({{0}}, {{}, 2})}, + {{"sscanf"}, TR::Prop({{0}}, {{}, 2})}, + {{"getc"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"getc_unlocked"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"getdelim"}, TR::Prop({{3}}, {{0}})}, + {{"getline"}, TR::Prop({{2}}, {{0}})}, + {{"getw"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"pread"}, TR::Prop({{0, 1, 2, 3}}, {{1, ReturnValueIndex}})}, + {{"read"}, TR::Prop({{0, 2}}, {{1, ReturnValueIndex}})}, + {{"strchr"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"strrchr"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"tolower"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{"toupper"}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {BI.getName(Builtin::BIstrncat)}}, + TR::Prop({{1, 2}}, {{0, ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {BI.getName(Builtin::BIstrlcpy)}}, + TR::Prop({{1, 2}}, {{0}})}, + {{CDF_MaybeBuiltin, {BI.getName(Builtin::BIstrlcat)}}, + TR::Prop({{1, 2}}, {{0}})}, + {{CDF_MaybeBuiltin, {"snprintf"}}, + TR::Prop({{1}, 3}, {{0, ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"sprintf"}}, + TR::Prop({{1}, 2}, {{0, ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"strcpy"}}, + TR::Prop({{1}}, {{0, ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"stpcpy"}}, + TR::Prop({{1}}, {{0, ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"strcat"}}, + TR::Prop({{1}}, {{0, ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"strdup"}}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"strdupa"}}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + {{CDF_MaybeBuiltin, {"wcsdup"}}, TR::Prop({{0}}, {{ReturnValueIndex}})}, + + // Sinks + {{"system"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"popen"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"execl"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"execle"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"execlp"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"execvp"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"execvP"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"execve"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{"dlopen"}, TR::Sink({{0}}, MsgSanitizeSystemArgs)}, + {{CDF_MaybeBuiltin, {"malloc"}}, TR::Sink({{0}}, MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {"calloc"}}, TR::Sink({{0}}, MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {"alloca"}}, TR::Sink({{0}}, MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {"memccpy"}}, TR::Sink({{3}}, MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {"realloc"}}, TR::Sink({{1}}, MsgTaintedBufferSize)}, + {{{"setproctitle"}}, TR::Sink({{0}, 1}, MsgUncontrolledFormatString)}, + {{{"setproctitle_fast"}}, + TR::Sink({{0}, 1}, MsgUncontrolledFormatString)}, + + // SinkProps + {{CDF_MaybeBuiltin, BI.getName(Builtin::BImemcpy)}, + TR::SinkProp({{2}}, {{1, 2}}, {{0, ReturnValueIndex}}, + MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {BI.getName(Builtin::BImemmove)}}, + TR::SinkProp({{2}}, {{1, 2}}, {{0, ReturnValueIndex}}, + MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {BI.getName(Builtin::BIstrncpy)}}, + TR::SinkProp({{2}}, {{1, 2}}, {{0, ReturnValueIndex}}, + MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {BI.getName(Builtin::BIstrndup)}}, + TR::SinkProp({{1}}, {{0, 1}}, {{ReturnValueIndex}}, + MsgTaintedBufferSize)}, + {{CDF_MaybeBuiltin, {"bcopy"}}, + TR::SinkProp({{2}}, {{0, 2}}, {{1}}, MsgTaintedBufferSize)}}; // `getenv` returns taint only in untrusted environments. - if (FData.FullName == "getenv") { - if (C.getAnalysisManager() - .getAnalyzerOptions() - .ShouldAssumeControlledEnvironment) - return {}; - return {{}, {ReturnValueIndex}}; + if (TR::UntrustedEnv(C)) { + // void setproctitle_init(int argc, char *argv[], char *envp[]) + GlobalCRules.push_back( + {{{"setproctitle_init"}}, TR::Sink({{2}}, MsgCustomSink)}); + GlobalCRules.push_back({{"getenv"}, TR::Source({{ReturnValueIndex}})}); } - assert(FData.FDecl); - - // Check if it's one of the memory setting/copying functions. - // This check is specialized but faster then calling isCLibraryFunction. - const FunctionDecl *FDecl = FData.FDecl; - unsigned BId = 0; - if ((BId = FDecl->getMemoryFunctionKind())) { - switch (BId) { - case Builtin::BImemcpy: - case Builtin::BImemmove: - case Builtin::BIstrncpy: - case Builtin::BIstrncat: - return {{1, 2}, {0, ReturnValueIndex}}; - case Builtin::BIstrlcpy: - case Builtin::BIstrlcat: - return {{1, 2}, {0}}; - case Builtin::BIstrndup: - return {{0, 1}, {ReturnValueIndex}}; - - default: - break; - } - } + StaticTaintRules.emplace(std::make_move_iterator(GlobalCRules.begin()), + std::make_move_iterator(GlobalCRules.end())); - // Process all other functions which could be defined as builtins. - if (Rule.isNull()) { - const auto OneOf = [FDecl](const auto &... Name) { - // FIXME: use fold expression in C++17 - using unused = int[]; - bool ret = false; - static_cast(unused{ - 0, (ret |= CheckerContext::isCLibraryFunction(FDecl, Name), 0)...}); - return ret; - }; - if (OneOf("snprintf")) - return {{1}, {0, ReturnValueIndex}, VariadicType::Src, 3}; - if (OneOf("sprintf")) - return {{1}, {0, ReturnValueIndex}, VariadicType::Src, 2}; - if (OneOf("strcpy", "stpcpy", "strcat")) - return {{1}, {0, ReturnValueIndex}}; - if (OneOf("bcopy")) - return {{0, 2}, {1}}; - if (OneOf("strdup", "strdupa", "wcsdup")) - return {{0}, {ReturnValueIndex}}; + // User-provided taint configuration. + CheckerManager *Mgr = C.getAnalysisManager().getCheckerManager(); + assert(Mgr); + GenericTaintRuleParser ConfigParser{*Mgr}; + std::string Option{"Config"}; + StringRef ConfigFile = + Mgr->getAnalyzerOptions().getCheckerStringOption(this, Option); + llvm::Optional Config = + getConfiguration(*Mgr, this, Option, ConfigFile); + if (!Config) { + // We don't have external taint config, no parsing required. + DynamicTaintRules = RuleLookupTy{}; + return; } - // Skipping the following functions, since they might be used for cleansing or - // smart memory copy: - // - memccpy - copying until hitting a special character. + GenericTaintRuleParser::RulesContTy Rules{ + ConfigParser.parseConfiguration(Option, std::move(Config.getValue()))}; - auto It = findFunctionInConfig(CustomPropagations, FData); - if (It != CustomPropagations.end()) - return It->second.second; - return {}; + DynamicTaintRules.emplace(std::make_move_iterator(Rules.begin()), + std::make_move_iterator(Rules.end())); } void GenericTaintChecker::checkPreCall(const CallEvent &Call, CheckerContext &C) const { - Optional FData = FunctionData::create(Call, C); - if (!FData) - return; - - // Check for taintedness related errors first: system call, uncontrolled - // format string, tainted buffer size. - if (checkPre(Call, *FData, C)) - return; - - // Marks the function's arguments and/or return value tainted if it present in - // the list. - if (addSourcesPre(Call, *FData, C)) - return; - - addFiltersPre(Call, *FData, C); + initTaintRules(C); + + // FIXME: this should be much simpler. + if (const auto *Rule = + Call.isGlobalCFunction() ? StaticTaintRules->lookup(Call) : nullptr) + Rule->process(*this, Call, C); + else if (const auto *Rule = DynamicTaintRules->lookup(Call)) + Rule->process(*this, Call, C); + + // FIXME: These edge cases are to be eliminated from here eventually. + // + // Additional check that is not supported by CallDescription. + // TODO: Make CallDescription be able to match attributes such as printf-like + // arguments. + checkUncontrolledFormatString(Call, C); + + // TODO: Modeling sockets should be done in a specific checker. + // Socket is a source, which taints the return value. + taintUnsafeSocketProtocol(Call, C); } void GenericTaintChecker::checkPostCall(const CallEvent &Call, CheckerContext &C) const { // Set the marked values as tainted. The return value only accessible from // checkPostStmt. - propagateFromPre(Call, C); -} - -void GenericTaintChecker::printState(raw_ostream &Out, ProgramStateRef State, - const char *NL, const char *Sep) const { - printTaint(State, Out, NL, Sep); -} - -bool GenericTaintChecker::addSourcesPre(const CallEvent &Call, - const FunctionData &FData, - CheckerContext &C) const { - // First, try generating a propagation rule for this function. - TaintPropagationRule Rule = TaintPropagationRule::getTaintPropagationRule( - this->CustomPropagations, FData, C); - if (!Rule.isNull()) { - ProgramStateRef State = Rule.process(Call, C); - if (State) { - C.addTransition(State); - return true; - } - } - return false; -} - -bool GenericTaintChecker::addFiltersPre(const CallEvent &Call, - const FunctionData &FData, - CheckerContext &C) const { - auto It = findFunctionInConfig(CustomFilters, FData); - if (It == CustomFilters.end()) - return false; - - ProgramStateRef State = C.getState(); - const auto &Value = It->second; - const ArgVector &Args = Value.second; - for (unsigned ArgNum : Args) { - if (ArgNum >= Call.getNumArgs()) - continue; - - const Expr *Arg = Call.getArgExpr(ArgNum); - Optional V = getPointeeOf(C, Arg); - if (V) - State = removeTaint(State, *V); - } - - if (State != C.getState()) { - C.addTransition(State); - return true; - } - return false; -} - -bool GenericTaintChecker::propagateFromPre(const CallEvent &Call, - CheckerContext &C) { ProgramStateRef State = C.getState(); // Depending on what was tainted at pre-visit, we determined a set of @@ -616,9 +689,9 @@ // stored in the state as TaintArgsOnPostVisit set. TaintArgsOnPostVisitTy TaintArgs = State->get(); if (TaintArgs.isEmpty()) - return false; + return; - for (unsigned ArgNum : TaintArgs) { + for (ArgIdxTy ArgNum : TaintArgs) { // Special handling for the tainted return value. if (ArgNum == ReturnValueIndex) { State = addTaint(State, Call.getReturnValue()); @@ -627,234 +700,147 @@ // The arguments are pointer arguments. The data they are pointing at is // tainted after the call. - if (Call.getNumArgs() < (ArgNum + 1)) - return false; - const Expr *Arg = Call.getArgExpr(ArgNum); - Optional V = getPointeeOf(C, Arg); - if (V) + if (auto V = getPointeeOf(C, Call.getArgSVal(ArgNum))) State = addTaint(State, *V); } // Clear up the taint info from the state. State = State->remove(); - - if (State != C.getState()) { - C.addTransition(State); - return true; - } - return false; -} - -bool GenericTaintChecker::checkPre(const CallEvent &Call, - const FunctionData &FData, - CheckerContext &C) const { - if (checkUncontrolledFormatString(Call, C)) - return true; - - if (checkSystemCall(Call, FData.Name, C)) - return true; - - if (checkTaintedBufferSize(Call, C)) - return true; - - return checkCustomSinks(Call, FData, C); + C.addTransition(State); } -Optional GenericTaintChecker::getPointeeOf(CheckerContext &C, - const Expr *Arg) { - ProgramStateRef State = C.getState(); - SVal AddrVal = C.getSVal(Arg->IgnoreParens()); - if (AddrVal.isUnknownOrUndef()) - return None; - - Optional AddrLoc = AddrVal.getAs(); - if (!AddrLoc) - return None; - - QualType ArgTy = Arg->getType().getCanonicalType(); - if (!ArgTy->isPointerType()) - return State->getSVal(*AddrLoc); - - QualType ValTy = ArgTy->getPointeeType(); - - // Do not dereference void pointers. Treat them as byte pointers instead. - // FIXME: we might want to consider more than just the first byte. - if (ValTy->isVoidType()) - ValTy = C.getASTContext().CharTy; - - return State->getSVal(*AddrLoc, ValTy); +void GenericTaintChecker::printState(raw_ostream &Out, ProgramStateRef State, + const char *NL, const char *Sep) const { + printTaint(State, Out, NL, Sep); } -ProgramStateRef -GenericTaintChecker::TaintPropagationRule::process(const CallEvent &Call, - CheckerContext &C) const { +void GenericTaintRule::process(const GenericTaintChecker &Checker, + const CallEvent &Call, CheckerContext &C) const { ProgramStateRef State = C.getState(); + const ArgIdxTy CallNumArgs = fromArgumentCount(Call.getNumArgs()); - // Check for taint in arguments. - bool IsTainted = true; - for (unsigned ArgNum : SrcArgs) { - if (ArgNum >= Call.getNumArgs()) - continue; - - if ((IsTainted = - isTaintedOrPointsToTainted(Call.getArgExpr(ArgNum), State, C))) - break; - } - - // Check for taint in variadic arguments. - if (!IsTainted && VariadicType::Src == VarType) { - // Check if any of the arguments is tainted - for (unsigned i = VariadicIndex; i < Call.getNumArgs(); ++i) { - if ((IsTainted = - isTaintedOrPointsToTainted(Call.getArgExpr(i), State, C))) - break; + /// Iterate every call argument, and get their corresponding Expr and SVal. + const auto ForEachCallArg = [&C, &Call, CallNumArgs](auto &&Fun) { + for (ArgIdxTy I = ReturnValueIndex; I < CallNumArgs; ++I) { + const Expr *E = GetArgExpr(I, Call); + Fun(I, E, C.getSVal(E)); } - } + }; - if (PropagationFunc) - IsTainted = PropagationFunc(IsTainted, Call, C); + /// Check for taint sinks. + ForEachCallArg([this, &Checker, &C, &State](ArgIdxTy I, const Expr *E, SVal) { + if (SinkArgs.contains(I) && isTaintedOrPointsToTainted(E, State, C)) + Checker.generateReportIfTainted(E, SinkMsg.getValueOr(MsgCustomSink), C); + }); + + /// Check for taint filters. + ForEachCallArg([this, &C, &State](ArgIdxTy I, const Expr *E, SVal S) { + if (FilterArgs.contains(I)) { + State = removeTaint(State, S); + if (auto P = getPointeeOf(C, S)) + State = removeTaint(State, *P); + } + }); + + /// Check for taint propagation sources. + /// A rule is relevant if PropSrcArgs is empty, or if any of its signified + /// args are tainted in context of the current CallEvent. + bool IsMatching = PropSrcArgs.isEmpty(); + ForEachCallArg( + [this, &C, &IsMatching, &State](ArgIdxTy I, const Expr *E, SVal) { + IsMatching = IsMatching || (PropSrcArgs.contains(I) && + isTaintedOrPointsToTainted(E, State, C)); + }); - if (!IsTainted) - return State; + if (!IsMatching) + return; - // Mark the arguments which should be tainted after the function returns. - for (unsigned ArgNum : DstArgs) { - // Should mark the return value? - if (ArgNum == ReturnValueIndex) { - State = State->add(ReturnValueIndex); - continue; - } + const auto WouldEscape = [](SVal V, QualType Ty) -> bool { + if (!V.getAs()) + return false; - if (ArgNum >= Call.getNumArgs()) - continue; + const bool IsNonConstRef = Ty->isReferenceType() && !Ty.isConstQualified(); + const bool IsNonConstPtr = + Ty->isPointerType() && !Ty->getPointeeType().isConstQualified(); - // Mark the given argument. - State = State->add(ArgNum); - } + return IsNonConstRef || IsNonConstPtr; + }; - // Mark all variadic arguments tainted if present. - if (VariadicType::Dst == VarType) { - // For all pointer and references that were passed in: - // If they are not pointing to const data, mark data as tainted. - // TODO: So far we are just going one level down; ideally we'd need to - // recurse here. - for (unsigned i = VariadicIndex; i < Call.getNumArgs(); ++i) { - const Expr *Arg = Call.getArgExpr(i); - // Process pointer argument. - const Type *ArgTy = Arg->getType().getTypePtr(); - QualType PType = ArgTy->getPointeeType(); - if ((!PType.isNull() && !PType.isConstQualified()) || - (ArgTy->isReferenceType() && !Arg->getType().isConstQualified())) { - State = State->add(i); - } - } - } + /// Propagate taint where it is necessary. + ForEachCallArg( + [this, &State, WouldEscape](ArgIdxTy I, const Expr *E, SVal V) { + if (PropDstArgs.contains(I)) + State = State->add(I); + + // TODO: We should traverse all reachable memory regions via the + // escaping parameter. Instead of doing that we simply mark only the + // referred memory region as tainted. + if (WouldEscape(V, E->getType())) + State = State->add(I); + }); - return State; + C.addTransition(State); } -// If argument 0(protocol domain) is network, the return value should get taint. -bool GenericTaintChecker::TaintPropagationRule::postSocket( - bool /*IsTainted*/, const CallEvent &Call, CheckerContext &C) { - SourceLocation DomLoc = Call.getArgExpr(0)->getExprLoc(); - StringRef DomName = C.getMacroNameOrSpelling(DomLoc); - // White list the internal communication protocols. - if (DomName.equals("AF_SYSTEM") || DomName.equals("AF_LOCAL") || - DomName.equals("AF_UNIX") || DomName.equals("AF_RESERVED_36")) - return false; - return true; +bool GenericTaintRule::UntrustedEnv(CheckerContext &C) { + return !C.getAnalysisManager() + .getAnalyzerOptions() + .ShouldAssumeControlledEnvironment; } -bool GenericTaintChecker::isStdin(const Expr *E, CheckerContext &C) { - ProgramStateRef State = C.getState(); - SVal Val = C.getSVal(E); - - // stdin is a pointer, so it would be a region. - const MemRegion *MemReg = Val.getAsRegion(); +bool GenericTaintChecker::generateReportIfTainted(const Expr *E, StringRef Msg, + CheckerContext &C) const { + assert(E); + Optional TaintedSVal{getTaintedPointeeOrPointer(C, C.getSVal(E))}; - // The region should be symbolic, we do not know it's value. - const auto *SymReg = dyn_cast_or_null(MemReg); - if (!SymReg) + if (!TaintedSVal) return false; - // Get it's symbol and find the declaration region it's pointing to. - const auto *Sm = dyn_cast(SymReg->getSymbol()); - if (!Sm) - return false; - const auto *DeclReg = dyn_cast_or_null(Sm->getRegion()); - if (!DeclReg) - return false; - - // This region corresponds to a declaration, find out if it's a global/extern - // variable named stdin with the proper type. - if (const auto *D = dyn_cast_or_null(DeclReg->getDecl())) { - D = D->getCanonicalDecl(); - if (D->getName().contains("stdin") && D->isExternC()) { - const auto *PtrTy = dyn_cast(D->getType().getTypePtr()); - if (PtrTy && PtrTy->getPointeeType().getCanonicalType() == - C.getASTContext().getFILEType().getCanonicalType()) - return true; - } + // Generate diagnostic. + if (ExplodedNode *N = C.generateNonFatalErrorNode()) { + auto report = std::make_unique(BT, Msg, N); + report->addRange(E->getSourceRange()); + report->addVisitor(std::make_unique(*TaintedSVal)); + C.emitReport(std::move(report)); + return true; } return false; } +/// TODO: remove checking for printf format attributes and socket whitelisting +/// from GenericTaintChecker, and that means the following functions: +/// getPrintfFormatArgumentNum, +/// GenericTaintChecker::checkUncontrolledFormatString, +/// GenericTaintChecker::taintUnsafeSocketProtocol + static bool getPrintfFormatArgumentNum(const CallEvent &Call, const CheckerContext &C, - unsigned &ArgNum) { + ArgIdxTy &ArgNum) { // Find if the function contains a format string argument. // Handles: fprintf, printf, sprintf, snprintf, vfprintf, vprintf, vsprintf, // vsnprintf, syslog, custom annotated functions. - const FunctionDecl *FDecl = Call.getDecl()->getAsFunction(); + const Decl *CallDecl = Call.getDecl(); + if (!CallDecl) + return false; + const FunctionDecl *FDecl = CallDecl->getAsFunction(); if (!FDecl) return false; + + const ArgIdxTy CallNumArgs = fromArgumentCount(Call.getNumArgs()); + for (const auto *Format : FDecl->specific_attrs()) { ArgNum = Format->getFormatIdx() - 1; - if ((Format->getType()->getName() == "printf") && - Call.getNumArgs() > ArgNum) + if ((Format->getType()->getName() == "printf") && CallNumArgs > ArgNum) return true; } - // Or if a function is named setproctitle (this is a heuristic). - if (C.getCalleeName(FDecl).contains("setproctitle")) { - ArgNum = 0; - return true; - } - - return false; -} - -bool GenericTaintChecker::generateReportIfTainted(const Expr *E, StringRef Msg, - CheckerContext &C) const { - assert(E); - - // Check for taint. - ProgramStateRef State = C.getState(); - Optional PointedToSVal = getPointeeOf(C, E); - SVal TaintedSVal; - if (PointedToSVal && isTainted(State, *PointedToSVal)) - TaintedSVal = *PointedToSVal; - else if (isTainted(State, E, C.getLocationContext())) - TaintedSVal = C.getSVal(E); - else - return false; - - // Generate diagnostic. - if (ExplodedNode *N = C.generateNonFatalErrorNode()) { - initBugType(); - auto report = std::make_unique(*BT, Msg, N); - report->addRange(E->getSourceRange()); - report->addVisitor(std::make_unique(TaintedSVal)); - C.emitReport(std::move(report)); - return true; - } return false; } bool GenericTaintChecker::checkUncontrolledFormatString( const CallEvent &Call, CheckerContext &C) const { // Check if the function contains a format string argument. - unsigned ArgNum = 0; + ArgIdxTy ArgNum = 0; if (!getPrintfFormatArgumentNum(Call, C, ArgNum)) return false; @@ -864,102 +850,32 @@ MsgUncontrolledFormatString, C); } -bool GenericTaintChecker::checkSystemCall(const CallEvent &Call, StringRef Name, - CheckerContext &C) const { - // TODO: It might make sense to run this check on demand. In some cases, - // we should check if the environment has been cleansed here. We also might - // need to know if the user was reset before these calls(seteuid). - unsigned ArgNum = llvm::StringSwitch(Name) - .Case("system", 0) - .Case("popen", 0) - .Case("execl", 0) - .Case("execle", 0) - .Case("execlp", 0) - .Case("execv", 0) - .Case("execvp", 0) - .Case("execvP", 0) - .Case("execve", 0) - .Case("dlopen", 0) - .Default(InvalidArgIndex); - - if (ArgNum == InvalidArgIndex || Call.getNumArgs() < (ArgNum + 1)) - return false; - - return generateReportIfTainted(Call.getArgExpr(ArgNum), MsgSanitizeSystemArgs, - C); -} - -// TODO: Should this check be a part of the CString checker? -// If yes, should taint be a global setting? -bool GenericTaintChecker::checkTaintedBufferSize(const CallEvent &Call, - CheckerContext &C) const { - const auto *FDecl = Call.getDecl()->getAsFunction(); - // If the function has a buffer size argument, set ArgNum. - unsigned ArgNum = InvalidArgIndex; - unsigned BId = 0; - if ((BId = FDecl->getMemoryFunctionKind())) { - switch (BId) { - case Builtin::BImemcpy: - case Builtin::BImemmove: - case Builtin::BIstrncpy: - ArgNum = 2; - break; - case Builtin::BIstrndup: - ArgNum = 1; - break; - default: - break; - } - } +void GenericTaintChecker::taintUnsafeSocketProtocol(const CallEvent &Call, + CheckerContext &C) const { + if (Call.getNumArgs() < 1) + return; + const IdentifierInfo *ID = Call.getCalleeIdentifier(); + if (!ID) + return; + if (!ID->getName().equals("socket")) + return; - if (ArgNum == InvalidArgIndex) { - using CCtx = CheckerContext; - if (CCtx::isCLibraryFunction(FDecl, "malloc") || - CCtx::isCLibraryFunction(FDecl, "calloc") || - CCtx::isCLibraryFunction(FDecl, "alloca")) - ArgNum = 0; - else if (CCtx::isCLibraryFunction(FDecl, "memccpy")) - ArgNum = 3; - else if (CCtx::isCLibraryFunction(FDecl, "realloc")) - ArgNum = 1; - else if (CCtx::isCLibraryFunction(FDecl, "bcopy")) - ArgNum = 2; - } + SourceLocation DomLoc = Call.getArgExpr(0)->getExprLoc(); + StringRef DomName = C.getMacroNameOrSpelling(DomLoc); + // Allow internal communication protocols. + bool SafeProtocol = DomName.equals("AF_SYSTEM") || + DomName.equals("AF_LOCAL") || DomName.equals("AF_UNIX") || + DomName.equals("AF_RESERVED_36"); + if (SafeProtocol) + return; - return ArgNum != InvalidArgIndex && Call.getNumArgs() > ArgNum && - generateReportIfTainted(Call.getArgExpr(ArgNum), MsgTaintedBufferSize, - C); + C.addTransition(C.getState()->add(ReturnValueIndex)); } -bool GenericTaintChecker::checkCustomSinks(const CallEvent &Call, - const FunctionData &FData, - CheckerContext &C) const { - auto It = findFunctionInConfig(CustomSinks, FData); - if (It == CustomSinks.end()) - return false; - - const auto &Value = It->second; - const GenericTaintChecker::ArgVector &Args = Value.second; - for (unsigned ArgNum : Args) { - if (ArgNum >= Call.getNumArgs()) - continue; - - if (generateReportIfTainted(Call.getArgExpr(ArgNum), MsgCustomSink, C)) - return true; - } - - return false; -} +/// Checker registration void ento::registerGenericTaintChecker(CheckerManager &Mgr) { - auto *Checker = Mgr.registerChecker(); - std::string Option{"Config"}; - StringRef ConfigFile = - Mgr.getAnalyzerOptions().getCheckerStringOption(Checker, Option); - llvm::Optional Config = - getConfiguration(Mgr, Checker, Option, ConfigFile); - if (Config) - Checker->parseConfiguration(Mgr, Option, std::move(Config.getValue())); + Mgr.registerChecker(); } bool ento::shouldRegisterGenericTaintChecker(const CheckerManager &mgr) {