/* * Cppcheck - A tool for static C/C++ code analysis * Copyright (C) 2007-2011 Daniel Marjamäki and Cppcheck team. * * This program is free software: you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation, either version 3 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program. If not, see . */ //--------------------------------------------------------------------------- // Buffer overrun.. //--------------------------------------------------------------------------- #include "checkbufferoverrun.h" #include "tokenize.h" #include "errorlogger.h" #include "mathlib.h" #include "symboldatabase.h" #include #include #include #include #include #include #include // <- assert #include // <- strtoul //--------------------------------------------------------------------------- // Register this check class (by creating a static instance of it) namespace { CheckBufferOverrun instance; } //--------------------------------------------------------------------------- static void makeArrayIndexOutOfBoundsError(std::ostream& oss, const CheckBufferOverrun::ArrayInfo &arrayInfo, const std::vector &index) { oss << "Array '" << arrayInfo.varname(); for (unsigned int i = 0; i < arrayInfo.num().size(); ++i) oss << "[" << arrayInfo.num(i) << "]"; oss << "' index "; if (index.size() == 1) oss << index[0]; else { oss << arrayInfo.varname(); for (unsigned int i = 0; i < index.size(); ++i) oss << "[" << index[i] << "]"; } oss << " out of bounds"; } void CheckBufferOverrun::arrayIndexOutOfBoundsError(const Token *tok, const ArrayInfo &arrayInfo, const std::vector &index) { std::ostringstream oss; makeArrayIndexOutOfBoundsError(oss, arrayInfo, index); reportError(tok, Severity::error, "arrayIndexOutOfBounds", oss.str()); } void CheckBufferOverrun::arrayIndexOutOfBoundsError(const std::list &callstack, const ArrayInfo &arrayInfo, const std::vector &index) { std::ostringstream oss; makeArrayIndexOutOfBoundsError(oss, arrayInfo, index); reportError(callstack, Severity::error, "arrayIndexOutOfBounds", oss.str()); } void CheckBufferOverrun::bufferOverrunError(const Token *tok, const std::string &varnames) { std::string v = varnames; while (v.find(" ") != std::string::npos) v.erase(v.find(" "), 1); std::string errmsg("Buffer access out-of-bounds"); if (!v.empty()) errmsg += ": " + v; reportError(tok, Severity::error, "bufferAccessOutOfBounds", errmsg); } void CheckBufferOverrun::possibleBufferOverrunError(const Token *tok, const std::string &src, const std::string &dst, bool cat) { if (cat) reportError(tok, Severity::warning, "possibleBufferAccessOutOfBounds", "Possible buffer overflow if strlen(" + src + ") is larger than sizeof(" + dst + ")-strlen(" + dst +").\n" "The source buffer is larger than the destination buffer so there is the potential for overflowing the destination buffer."); else reportError(tok, Severity::warning, "possibleBufferAccessOutOfBounds", "Possible buffer overflow if strlen(" + src + ") is larger than or equal to sizeof(" + dst + ").\n" "The source buffer is larger than the destination buffer so there is the potential for overflowing the destination buffer."); } void CheckBufferOverrun::possibleReadlinkBufferOverrunError(const Token* tok, const std::string &funcname, const std::string &varname) { const std::string errmsg = funcname + "() might return the full size of '" + varname + "'. Lower the supplied size by one.\n" + funcname + "() might return the full size of '" + varname + "'. Lower the supplied size by one. " "If a " + varname + "[len] = '\\0'; statement follows, it will overrun the buffer."; reportInconclusiveError(tok, Severity::warning, "possibleReadlinkBufferOverrun", errmsg); } void CheckBufferOverrun::strncatUsageError(const Token *tok) { if (_settings && !_settings->isEnabled("style")) return; reportError(tok, Severity::warning, "strncatUsage", "Dangerous usage of strncat - 3rd parameter is the maximum number of characters to append.\n" "strncat appends at max its 3rd parameter's amount of characters. The safe way to use " "strncat is to calculate remaining space in the buffer and use it as 3rd parameter."); } void CheckBufferOverrun::outOfBoundsError(const Token *tok, const std::string &what, const bool show_size_info, const MathLib::bigint &supplied_size, const MathLib::bigint &actual_size) { std::ostringstream oss; oss << what << " is out of bounds"; if (show_size_info) oss << ": Supplied size " << supplied_size << " is larger than actual size of " << actual_size; reportError(tok, Severity::error, "outOfBounds", oss.str()); } void CheckBufferOverrun::pointerOutOfBoundsError(const Token *tok, const std::string &object) { reportError(tok, Severity::portability, "pointerOutOfBounds", "Undefined behaviour: pointer arithmetic result does not point into or just past the end of the " + object + "\n" "Undefined behaviour: Using pointer arithmetic so that the result does not point into or just past the end of the same object. Further information: https://www.securecoding.cert.org/confluence/display/seccode/ARR30-C.+Do+not+form+or+use+out+of+bounds+pointers+or+array+subscripts"); } void CheckBufferOverrun::sizeArgumentAsCharError(const Token *tok) { if (_settings && !_settings->isEnabled("style")) return; reportError(tok, Severity::warning, "sizeArgumentAsChar", "The size argument is given as a char constant"); } void CheckBufferOverrun::terminateStrncpyError(const Token *tok, const std::string &varname) { reportError(tok, Severity::warning, "terminateStrncpy", "The buffer '" + varname + "' may not be zero-terminated after the call to strncpy().\n" "The use of strncpy() usually indicates that the programmer wants to ensure " "the buffer is zero-terminated after the call. However if the (buffer) size given for " "the strncpy() call matches the actual buffer size strncpy() does not add the " "zero at the end of the buffer. This may cause bugs later in the code if " "the code assumes buffer is zero-terminated."); } void CheckBufferOverrun::cmdLineArgsError(const Token *tok) { reportError(tok, Severity::error, "insecureCmdLineArgs", "Buffer overrun possible for long cmd-line args"); } void CheckBufferOverrun::bufferNotZeroTerminatedError(const Token *tok, const std::string &varname, const std::string &function) { const std::string errmsg = "The buffer '" + varname + "' is not zero-terminated after the call to " + function + "().\n" "The buffer '" + varname + "' is not zero-terminated after the call to " + function + "(). " "This will cause bugs later in the code if the code assumes the buffer is zero-terminated."; reportInconclusiveError(tok, Severity::warning, "bufferNotZeroTerminated", errmsg); } //--------------------------------------------------------------------------- //--------------------------------------------------------------------------- // Check array usage.. //--------------------------------------------------------------------------- /** * @brief This is a helper class to be used with std::find_if */ class TokenStrEquals { public: /** * @param str Token::str() is compared against this. */ explicit TokenStrEquals(const std::string &str) : value(str) { } /** * Called automatically by std::find_if * @param tok Token inside the list */ bool operator()(const Token *tok) const { return value == tok->str(); } private: TokenStrEquals& operator=(const TokenStrEquals&); // disallow assignments const std::string value; }; /** * bailout if variable is used inside if/else/switch block or if there is "break" * @param tok token for "if" or "switch" * @param varid variable id * @return is bailout recommended? */ static bool bailoutIfSwitch(const Token *tok, const unsigned int varid) { // Used later to check if the body belongs to a "if" bool is_if = tok->str() == "if"; // Count { and } unsigned int indentlevel = 0; for (; tok; tok = tok->next()) { if (tok->str() == "{") indentlevel++; else if (tok->str() == "}") { // scan the else-block if (indentlevel == 1 && Token::simpleMatch(tok, "} else {")) { --indentlevel; continue; } else if (indentlevel <= 1) { break; } --indentlevel; } // If scanning a "if" block then bailout for "break" else if (is_if && tok->str() == "break") return true; // bailout for "return" else if (tok->str() == "return") return true; // bailout if varid is found else if (tok->varId() == varid) return true; } // No bailout stuff found => return false return false; } /** * Parse for loop initialization statement. Look for a counter variable * \param tok [in] first token inside the parentheses * \param varid [out] varid of counter variable * \param init_value [out] init value of counter variable * \return success => pointer to the for loop condition. fail => 0 */ static const Token *for_init(const Token *tok, unsigned int &varid, std::string &init_value) { if (Token::Match(tok, "%var% = %any% ;")) { if (tok->tokAt(2)->isNumber()) { init_value = tok->strAt(2); } varid = tok->varId(); tok = tok->tokAt(4); } else if (Token::Match(tok, "%type% %var% = %any% ;")) { if (tok->tokAt(3)->isNumber()) { init_value = tok->strAt(3); } varid = tok->next()->varId(); tok = tok->tokAt(5); } else if (Token::Match(tok, "%type% %type% %var% = %any% ;")) { if (tok->tokAt(4)->isNumber()) { init_value = tok->strAt(4); } varid = tok->tokAt(2)->varId(); tok = tok->tokAt(6); } else return 0; return tok; } /** Parse for condition */ static bool for_condition(const Token * const tok2, unsigned int varid, std::string &min_value, std::string &max_value, std::string &strindex, bool &maxMinFlipped) { if (Token::Match(tok2, "%varid% < %num% ;", varid) || Token::Match(tok2, "%varid% != %num% ; ++ %varid%", varid) || Token::Match(tok2, "%varid% != %num% ; %varid% ++", varid)) { maxMinFlipped = false; const MathLib::bigint value = MathLib::toLongNumber(tok2->strAt(2)); max_value = MathLib::toString(value - 1); } else if (Token::Match(tok2, "%varid% <= %num% ;", varid)) { maxMinFlipped = false; max_value = tok2->strAt(2); } else if (Token::Match(tok2, "%num% < %varid% ;", varid) || Token::Match(tok2, "%num% != %varid% ; ++ %varid%", varid) || Token::Match(tok2, "%num% != %varid% ; %varid% ++", varid)) { maxMinFlipped = true; const MathLib::bigint value = MathLib::toLongNumber(tok2->str()); max_value = min_value; min_value = MathLib::toString(value + 1); } else if (Token::Match(tok2, "%num% <= %varid% ;", varid)) { maxMinFlipped = true; max_value = min_value; min_value = tok2->str(); } else { return false; } strindex = tok2->isName() ? tok2->str() : tok2->strAt(2); return true; } /** * calculate maximum value of loop variable * @param stepvalue token that contains the step value * @param min_value the minimum value of loop variable * @param max_value maximum value of the loop variable */ static bool for_maxvalue(const Token * const stepvalue, const std::string &min_value, std::string &max_value) { if (!MathLib::isInt(stepvalue->str())) return false; // We have for example code: "for(i=2;i<22;i+=6) // We can calculate that max value for i is 20, not 21 // 21-2 = 19 // 19/6 = 3 // 6*3+2 = 20 const MathLib::bigint num = MathLib::toLongNumber(stepvalue->str()); MathLib::bigint max = MathLib::toLongNumber(max_value); const MathLib::bigint min = MathLib::toLongNumber(min_value); max = ((max - min) / num) * num + min; max_value = MathLib::toString(max); return true; } /** * Parse the third sub-statement in for head * \param tok first token * \param varid variable id of counter * \param min_value min value of counter * \param max_value max value of counter * \param maxMinFlipped counting from max to min */ static bool for3(const Token * const tok, unsigned int varid, std::string &min_value, std::string &max_value, const bool maxMinFlipped) { assert(tok != 0); if (Token::Match(tok, "%varid% += %num% )", varid) || Token::Match(tok, "%varid% = %num% + %varid% )", varid)) { if (!for_maxvalue(tok->tokAt(2), min_value, max_value)) return false; } else if (Token::Match(tok, "%varid% = %varid% + %num% )", varid)) { if (!for_maxvalue(tok->tokAt(4), min_value, max_value)) return false; } else if (Token::Match(tok, "%varid% -= %num% )", varid) || Token::Match(tok, "%varid% = %num% - %varid% )", varid)) { if (!for_maxvalue(tok->tokAt(2), min_value, max_value)) return false; } else if (Token::Match(tok, "%varid% = %varid% - %num% )", varid)) { if (!for_maxvalue(tok->tokAt(4), min_value, max_value)) return false; } else if (Token::Match(tok, "--| %varid% --| )", varid)) { if (!maxMinFlipped && MathLib::toLongNumber(min_value) < MathLib::toLongNumber(max_value)) { // Code relies on the fact that integer will overflow: // for (unsigned int i = 3; i < 5; --i) // Set min value in this case to zero. max_value = min_value; min_value = "0"; } } else if (! Token::Match(tok, "++| %varid% ++| )", varid)) { return false; } return true; } /** * Check is the counter variable increased elsewhere inside the loop or used * for anything else except reading * \param tok1 first token of for-body * \param varid counter variable id * \return bailout needed => true */ static bool for_bailout(const Token * const tok1, unsigned int varid) { for (const Token *loopTok = tok1; loopTok && loopTok != tok1->link(); loopTok = loopTok->next()) { if (loopTok->varId() == varid) { // Counter variable used inside loop if (Token::Match(loopTok->next(), "+=|-=|++|--|=") || Token::Match(loopTok->previous(), "++|--")) { return true; } } } return false; } void CheckBufferOverrun::parse_for_body(const Token *tok2, const ArrayInfo &arrayInfo, const std::string &strindex, bool condition_out_of_bounds, unsigned int counter_varid, const std::string &min_counter_value, const std::string &max_counter_value) { const std::string pattern((arrayInfo.varid() ? std::string("%varid%") : arrayInfo.varname()) + " [ " + strindex + " ]"); // count { and } for tok2 int indentlevel2 = 0; for (; tok2; tok2 = tok2->next()) { if (tok2->str() == ";" && indentlevel2 == 0) break; if (tok2->str() == "{") ++indentlevel2; if (tok2->str() == "}") { --indentlevel2; if (indentlevel2 <= 0) break; } // TODO: try to reduce false negatives. This is just a quick fix // for TestBufferOverrun::array_index_for_question if (tok2->str() == "?") break; if (Token::simpleMatch(tok2, "for (") && Token::simpleMatch(tok2->next()->link(), ") {")) { const Token *endpar = tok2->next()->link(); const Token *startbody = endpar->next(); const Token *endbody = startbody->link(); tok2 = endbody; continue; } if (Token::Match(tok2, "if|switch")) { if (bailoutIfSwitch(tok2, arrayInfo.varid())) break; } if (condition_out_of_bounds && Token::Match(tok2, pattern.c_str(), arrayInfo.varid())) { bufferOverrunError(tok2, arrayInfo.varname()); break; } else if (arrayInfo.varid() && counter_varid > 0 && !min_counter_value.empty() && !max_counter_value.empty()) { // Is the loop variable used to calculate the array index? // In this scope it is determined if such calculated // array indexes are out of bounds. // Only the minimum and maximum results of the calculation is // determined // Minimum calculated array index int min_index = 0; // Maximum calculated array index int max_index = 0; if (Token::Match(tok2, "%varid% [ %var% +|-|*|/ %num% ]", arrayInfo.varid()) && tok2->tokAt(2)->varId() == counter_varid) { // operator: +-*/ const char action = tok2->strAt(3)[0]; // second operator const std::string &second(tok2->strAt(4)); //printf("min_index: %s %c %s\n", min_counter_value.c_str(), action, second.c_str()); //printf("max_index: %s %c %s\n", max_counter_value.c_str(), action, second.c_str()); min_index = std::atoi(MathLib::calculate(min_counter_value, second, action, _tokenizer).c_str()); max_index = std::atoi(MathLib::calculate(max_counter_value, second, action, _tokenizer).c_str()); } else if (Token::Match(tok2, "%varid% [ %num% +|-|*|/ %var% ]", arrayInfo.varid()) && tok2->tokAt(4)->varId() == counter_varid) { // operator: +-*/ const char action = tok2->strAt(3)[0]; // first operand const std::string &first(tok2->strAt(2)); //printf("min_index: %s %c %s\n", first.c_str(), action, min_counter_value.c_str()); //printf("max_index: %s %c %s\n", first.c_str(), action, max_counter_value.c_str()); min_index = std::atoi(MathLib::calculate(first, min_counter_value, action, _tokenizer).c_str()); max_index = std::atoi(MathLib::calculate(first, max_counter_value, action, _tokenizer).c_str()); } //printf("min_index = %d, max_index = %d, size = %d\n", min_index, max_index, size); if (min_index < 0 || max_index < 0) { std::vector indexes; indexes.push_back(std::min(min_index, max_index)); arrayIndexOutOfBoundsError(tok2, arrayInfo, indexes); } // skip 0 length arrays if (arrayInfo.num(0) && (min_index >= arrayInfo.num(0) || max_index >= arrayInfo.num(0))) { std::vector indexes; indexes.push_back(std::max(min_index, max_index)); arrayIndexOutOfBoundsError(tok2, arrayInfo, indexes); } } } } void CheckBufferOverrun::checkFunctionParameter(const Token &tok, unsigned int par, const ArrayInfo &arrayInfo) { // total_size : which parameter in function call takes the total size? std::map total_size; total_size["fgets"] = 2; // The second argument for fgets can't exceed the total size of the array total_size["memcmp"] = 3; total_size["memcpy"] = 3; total_size["memmove"] = 3; total_size["memchr"] = 3; if (par == 1) { // reading from array // if it is zero terminated properly there won't be buffer overruns total_size["strncat"] = 3; total_size["strncpy"] = 3; total_size["memset"] = 3; total_size["fread"] = 1001; // parameter 2 * parameter 3 total_size["fwrite"] = 1001; // parameter 2 * parameter 3 } else if (par == 2) { total_size["read"] = 3; total_size["pread"] = 3; total_size["write"] = 3; total_size["recv"] = 3; total_size["recvfrom"] = 3; total_size["send"] = 3; total_size["sendto"] = 3; } std::map::const_iterator it = total_size.find(tok.str()); if (it != total_size.end()) { if (arrayInfo.element_size() == 0) return; // arg : the index of the "wanted" argument in the function call. unsigned int arg = it->second; // Parse function call. When a ',' is seen, arg is decremented. // if arg becomes 1 then the current function parameter is the wanted parameter. // if arg becomes 1000 then multiply current and next argument. for (const Token *tok2 = tok.tokAt(2); tok2; tok2 = tok2->next()) { if (tok2->str() == "(") { tok2 = tok2->link(); continue; } if (tok2->str() == ")") break; if (tok2->str() == ",") { --arg; if (arg == 1) { if (Token::Match(tok2, ", %num% ,|)")) { const MathLib::bigint sz = MathLib::toLongNumber(tok2->strAt(1)); MathLib::bigint elements = 1; for (unsigned int i = 0; i < arrayInfo.num().size(); ++i) elements *= arrayInfo.num(i); if (sz < 0 || sz > int(elements * arrayInfo.element_size())) { bufferOverrunError(&tok, arrayInfo.varname()); } } else if (Token::Match(tok2, ", %any% ,|)") && tok2->next()->str()[0] == '\'') { sizeArgumentAsCharError(tok2->next()); } break; } if (arg == 1000) { // special code. This parameter multiplied with the next must not exceed total_size if (Token::Match(tok2, ", %num% , %num% ,|)")) { const MathLib::bigint sz = MathLib::toLongNumber(MathLib::multiply(tok2->strAt(1), tok2->strAt(3))); MathLib::bigint elements = 1; for (unsigned int i = 0; i < arrayInfo.num().size(); ++i) elements *= arrayInfo.num(i); if (sz < 0 || sz > int(elements * arrayInfo.element_size())) { bufferOverrunError(&tok, arrayInfo.varname()); } } break; } } } } // Calling a user function? // only 1-dimensional arrays can be checked currently else if (arrayInfo.num().size() == 1) { const Token *ftok = _tokenizer->getFunctionTokenByName(tok.str().c_str()); if (Token::Match(ftok, "%var% (") && Token::Match(ftok->next()->link(), ") const| {")) { // Get varid for the corresponding parameter.. unsigned int parameter = 1; unsigned int parameterVarId = 0; for (const Token *ftok2 = ftok->tokAt(2); ftok2; ftok2 = ftok2->next()) { if (ftok2->str() == ",") { if (parameter >= par) break; ++parameter; } else if (ftok2->str() == ")") break; else if (parameter == par && Token::Match(ftok2, "%var% ,|)|[")) { // check type.. const Token *type = ftok2->previous(); while (Token::Match(type, "*|const")) type = type->previous(); if (type && _tokenizer->sizeOfType(type) == arrayInfo.element_size()) parameterVarId = ftok2->varId(); } } // No parameterVarId => bail out if (parameterVarId == 0) return; // Step into the function scope.. ftok = ftok->next()->link(); if (!Token::Match(ftok, ") const| {")) return; ftok = Token::findsimplematch(ftok, "{"); ftok = ftok->next(); // Check the parameter usage in the function scope.. for (; ftok; ftok = ftok->next()) { if (Token::Match(ftok, "if|for|while (")) { // bailout if there is buffer usage.. if (bailoutIfSwitch(ftok, parameterVarId)) { break; } // no bailout is needed. skip the if-block else { // goto end of if block.. ftok = ftok->next()->link()->next()->link(); if (Token::simpleMatch(ftok, "} else {")) ftok = ftok->linkAt(2); if (!ftok) break; continue; } } if (ftok->str() == "}") break; if (ftok->varId() == parameterVarId) { if (Token::Match(ftok->previous(), "-- %var%") || Token::Match(ftok, "%var% --")) break; if (Token::Match(ftok->previous(), "=|;|{|}|%op% %var% [ %num% ]")) { const MathLib::bigint index = MathLib::toLongNumber(ftok->strAt(2)); if (index >= 0 && arrayInfo.num(0) > 0 && index >= arrayInfo.num(0)) { std::list callstack; callstack.push_back(&tok); callstack.push_back(ftok); std::vector indexes; indexes.push_back(index); arrayIndexOutOfBoundsError(callstack, arrayInfo, indexes); } } } } } } } void CheckBufferOverrun::checkFunctionCall(const Token *tok, const ArrayInfo &arrayInfo) { // 1st parameter.. if (Token::Match(tok->tokAt(2), "%varid% ,|)", arrayInfo.varid())) checkFunctionParameter(*tok, 1, arrayInfo); else if (Token::Match(tok->tokAt(2), "%varid% + %num% ,|)", arrayInfo.varid())) { const ArrayInfo ai(arrayInfo.limit(MathLib::toLongNumber(tok->strAt(4)))); checkFunctionParameter(*tok, 1, ai); } // goto 2nd parameter and check it.. for (const Token *tok2 = tok->tokAt(2); tok2; tok2 = tok2->next()) { if (tok2->str() == "(") { tok2 = tok2->link(); continue; } if (tok2->str() == ";" || tok2->str() == ")") break; if (tok2->str() == ",") { if (Token::Match(tok2, ", %varid% ,|)", arrayInfo.varid())) checkFunctionParameter(*tok, 2, arrayInfo); else if (Token::Match(tok2, ", %varid% + %num% ,|)", arrayInfo.varid())) { const ArrayInfo ai(arrayInfo.limit(MathLib::toLongNumber(tok2->strAt(3)))); checkFunctionParameter(*tok, 2, ai); } break; } } } void CheckBufferOverrun::checkScopeForBody(const Token *tok, const ArrayInfo &arrayInfo, bool &bailout) { bailout = false; const Token *tok2 = tok->tokAt(2); const MathLib::bigint size = arrayInfo.num(0); // Check if there is a break in the body.. { const Token *bodyStart = tok->next()->link()->next(); const Token *bodyEnd = bodyStart->link(); if (Token::findsimplematch(bodyStart, "break ;", bodyEnd)) return; } unsigned int counter_varid = 0; std::string min_counter_value; std::string max_counter_value; tok2 = for_init(tok2, counter_varid, min_counter_value); if (tok2 == 0 || counter_varid == 0) return; bool maxMinFlipped = false; std::string strindex; if (!for_condition(tok2, counter_varid, min_counter_value, max_counter_value, strindex, maxMinFlipped)) return; // Get index variable and stopsize. bool condition_out_of_bounds = bool(size > 0); if (MathLib::toLongNumber(max_counter_value) < size) condition_out_of_bounds = false; if (!for3(tok2->tokAt(4), counter_varid, min_counter_value, max_counter_value, maxMinFlipped)) return; if (Token::Match(tok2->tokAt(4), "%var% =|+=|-=") && MathLib::toLongNumber(max_counter_value) <= size) condition_out_of_bounds = false; // Goto the end parenthesis of the for-statement: "for (x; y; z)" .. tok2 = tok->next()->link(); if (!tok2 || !tok2->tokAt(5)) { bailout = true; return; } // Check is the counter variable increased elsewhere inside the loop or used // for anything else except reading if (for_bailout(tok2->next(), counter_varid)) { bailout = true; return; } parse_for_body(tok2->next(), arrayInfo, strindex, condition_out_of_bounds, counter_varid, min_counter_value, max_counter_value); } void CheckBufferOverrun::checkScope(const Token *tok, const std::vector &varname, const ArrayInfo &arrayInfo) { const MathLib::bigint size = arrayInfo.num(0); const MathLib::bigint total_size = arrayInfo.element_size() * arrayInfo.num(0); unsigned int varid = arrayInfo.varid(); std::string varnames; for (unsigned int i = 0; i < varname.size(); ++i) varnames += (i == 0 ? "" : " . ") + varname[i]; const unsigned char varc(static_cast(varname.empty() ? 0U : (varname.size() - 1) * 2U)); if (Token::simpleMatch(tok, "return")) { tok = tok->next(); if (!tok) return; } // Array index.. if ((varid > 0 && Token::Match(tok, "%varid% [ %num% ]", varid)) || (varid == 0 && Token::Match(tok, (varnames + " [ %num% ]").c_str()))) { const MathLib::bigint index = MathLib::toLongNumber(tok->strAt(2 + varc)); if (index >= size) { std::vector indexes; indexes.push_back(index); arrayIndexOutOfBoundsError(tok->tokAt(varc), arrayInfo, indexes); } } // If the result of pointer arithmetic means that the pointer is // out of bounds then this flag will be set. bool pointerIsOutOfBounds = false; // Count { and } for tok int indentlevel = 0; for (; tok; tok = tok->next()) { if (tok->str() == "{") { ++indentlevel; } else if (tok->str() == "}") { --indentlevel; if (indentlevel < 0) return; } if (varid != 0 && Token::Match(tok, "%varid% = new|malloc|realloc", varid)) { // Abort break; } // reassign buffer if (varid > 0 && Token::Match(tok, "[;{}] %varid% =", varid)) { // using varid .. bailout if (!Token::Match(tok->tokAt(3), "%varid%", varid)) break; pointerIsOutOfBounds = false; } // Array index.. if ((varid > 0 && ((tok->str() == "return" || (!tok->isName() && !Token::Match(tok, "[.&]"))) && Token::Match(tok->next(), "%varid% [ %num% ]", varid))) || (varid == 0 && ((tok->str() == "return" || (!tok->isName() && !Token::Match(tok, "[.&]"))) && Token::Match(tok->next(), (varnames + " [ %num% ]").c_str())))) { std::vector indexes; const Token *tok2 = tok->tokAt(2 + varc); for (; Token::Match(tok2, "[ %num% ]"); tok2 = tok2->tokAt(3)) { const MathLib::bigint index = MathLib::toLongNumber(tok2->strAt(1)); indexes.push_back(index); } if (indexes.size() == arrayInfo.num().size()) { // Check if the indexes point outside the whole array.. // char a[10][10]; // a[0][20] <-- ok. // a[9][20] <-- error. // total number of elements of array.. MathLib::bigint totalElements = 1; // total index.. MathLib::bigint totalIndex = 0; // calculate the totalElements and totalIndex.. for (unsigned int i = 0; i < indexes.size(); ++i) { std::size_t ri = indexes.size() - 1 - i; totalIndex += indexes[ri] * totalElements; totalElements *= arrayInfo.num(ri); } // totalElements == 0 => Unknown size if (totalElements == 0) continue; const Token *tok3 = tok->previous(); while (tok3 && Token::Match(tok3->previous(), "%var% .")) tok3 = tok3->tokAt(-2); // just taking the address? const bool addr(tok3 && (tok3->str() == "&" || Token::simpleMatch(tok3->previous(), "& ("))); // taking address of 1 past end? if (addr && totalIndex == totalElements) continue; // Is totalIndex in bounds? if (totalIndex > totalElements || totalIndex < 0) { arrayIndexOutOfBoundsError(tok->tokAt(1 + varc), arrayInfo, indexes); } // Is any array index out of bounds? else { // check each index for overflow for (unsigned int i = 0; i < indexes.size(); ++i) { if (indexes[i] >= arrayInfo.num(i)) { // The access is still within the memory range for the array // so it may be intentional. if (_settings->inconclusive) { arrayIndexOutOfBoundsError(tok->tokAt(1 + varc), arrayInfo, indexes); break; // only warn about the first one } } } } } tok = tok2; continue; } // memset, memcmp, memcpy, strncpy, fgets.. if (varid == 0 && size > 0) { if (Token::Match(tok, ("%var% ( " + varnames + " ,").c_str())) checkFunctionParameter(*tok, 1, arrayInfo); if (Token::Match(tok, ("%var% ( %var% , " + varnames + " ,").c_str())) checkFunctionParameter(*tok, 2, arrayInfo); } // Loop.. if (Token::simpleMatch(tok, "for (")) { const ArrayInfo arrayInfo1(varid, varnames, (unsigned int)size, (unsigned int)total_size); bool bailout = false; checkScopeForBody(tok, arrayInfo1, bailout); if (bailout) break; continue; } // Writing data into array.. if ((varid > 0 && Token::Match(tok, "strcpy|strcat ( %varid% , %str% )", varid)) || (varid == 0 && Token::Match(tok, ("strcpy|strcat ( " + varnames + " , %str% )").c_str()))) { const std::size_t len = Token::getStrLength(tok->tokAt(varc + 4)); if (total_size > 0 && len >= (unsigned int)total_size) { bufferOverrunError(tok, varid > 0 ? std::string("") : varnames); continue; } } else if ((varid > 0 && Token::Match(tok, "strcpy|strcat ( %varid% , %var% )", varid)) || (varid == 0 && Token::Match(tok, ("strcpy|strcat ( " + varnames + " , %var% )").c_str()))) { const Variable *var = _tokenizer->getSymbolDatabase()->getVariableFromVarId(tok->tokAt(4)->varId()); if (var && var->isArray() && var->dimensions().size() == 1) { const std::size_t len = (std::size_t)var->dimension(0); if (total_size > 0 && len > (unsigned int)total_size) { if (_settings->inconclusive) possibleBufferOverrunError(tok, tok->strAt(4), tok->strAt(2), tok->str() == "strcat"); continue; } } } // Detect few strcat() calls const std::string strcatPattern = varid > 0 ? std::string("strcat ( %varid% , %str% ) ;") : ("strcat ( " + varnames + " , %str% ) ;"); if (Token::Match(tok, strcatPattern.c_str(), varid)) { size_t charactersAppend = 0; const Token *tok2 = tok; while (tok2 && Token::Match(tok2, strcatPattern.c_str(), varid)) { charactersAppend += Token::getStrLength(tok2->tokAt(4 + varc)); if (charactersAppend >= static_cast(total_size)) { bufferOverrunError(tok2); break; } tok2 = tok2->tokAt(7 + varc); } } // sprintf.. // TODO: change total_size to an unsigned value and remove the "&& total_size > 0" check. const std::string sprintfPattern = varid > 0 ? std::string("sprintf ( %varid% , %str% [,)]") : ("sprintf ( " + varnames + " , %str% [,)]"); if (Token::Match(tok, sprintfPattern.c_str(), varid) && total_size > 0) { checkSprintfCall(tok, static_cast(total_size)); } // snprintf.. const std::string snprintfPattern = varid > 0 ? std::string("snprintf ( %varid% , %num% ,") : ("snprintf ( " + varnames + " , %num% ,"); if (Token::Match(tok, snprintfPattern.c_str(), varid)) { const MathLib::bigint n = MathLib::toLongNumber(tok->strAt(4 + varc)); if (n > total_size) outOfBoundsError(tok->tokAt(4 + varc), "snprintf size", true, n, total_size); } // Check function call.. if (Token::Match(tok, "%var% (") && total_size > 0) { // No varid => function calls are not handled if (varid == 0) continue; const ArrayInfo arrayInfo1(varid, varnames, total_size / size, size); checkFunctionCall(tok, arrayInfo1); } // undefined behaviour: result of pointer arithmetic is out of bounds if (varid && Token::Match(tok, "= %varid% + %num% ;", varid)) { const MathLib::bigint index = MathLib::toLongNumber(tok->strAt(3)); if (index > size && _settings->isEnabled("portability")) pointerOutOfBoundsError(tok->next(), "buffer"); if (index >= size && Token::Match(tok->tokAt(-2), "[;{}] %varid% =", varid)) pointerIsOutOfBounds = true; } if (pointerIsOutOfBounds && Token::Match(tok, "[;{}=] * %varid% [;=]", varid)) { outOfBoundsError(tok->tokAt(2), tok->strAt(2), false, 0, 0); } } } void CheckBufferOverrun::checkScope(const Token *tok, const ArrayInfo &arrayInfo) { const MathLib::bigint total_size = arrayInfo.num(0) * arrayInfo.element_size(); const Token *scope_begin = tok->previous(); assert(scope_begin != 0); // Count { and } for tok unsigned int indentlevel = 0; for (; tok; tok = tok->next()) { if (tok->str() == "{") { ++indentlevel; } else if (tok->str() == "}") { if (indentlevel == 0) return; --indentlevel; } // Skip array declarations else if (Token::Match(tok, "[;{}] %type% *| %var% [") && tok->strAt(1) != "return") { tok = tok->tokAt(3); continue; } else if (Token::Match(tok, "%varid% [ %num% ]", arrayInfo.varid())) { std::vector indexes; for (const Token *tok2 = tok->next(); Token::Match(tok2, "[ %num% ]"); tok2 = tok2->tokAt(3)) { const MathLib::bigint index = MathLib::toLongNumber(tok2->strAt(1)); if (index < 0) { indexes.clear(); break; } indexes.push_back(index); } if (indexes.size() == arrayInfo.num().size()) { // Check if the indexes point outside the whole array.. // char a[10][10]; // a[0][20] <-- ok. // a[9][20] <-- error. // total number of elements of array.. MathLib::bigint totalElements = 1; // total index.. MathLib::bigint totalIndex = 0; // calculate the totalElements and totalIndex.. for (unsigned int i = 0; i < indexes.size(); ++i) { std::size_t ri = indexes.size() - 1 - i; totalIndex += indexes[ri] * totalElements; totalElements *= arrayInfo.num(ri); } // totalElements == 0 => Unknown size if (totalElements == 0) continue; const Token *tok2 = tok->previous(); while (tok2 && Token::Match(tok2->previous(), "%var% .")) tok2 = tok2->tokAt(-2); // just taking the address? const bool addr(tok2 && (tok2->str() == "&" || Token::simpleMatch(tok2->previous(), "& ("))); // taking address of 1 past end? if (addr && totalIndex == totalElements) continue; // Is totalIndex in bounds? if (totalIndex >= totalElements) { arrayIndexOutOfBoundsError(tok, arrayInfo, indexes); } // Is any array index out of bounds? else { // check each index for overflow for (unsigned int i = 0; i < indexes.size(); ++i) { if (indexes[i] >= arrayInfo.num(i)) { // The access is still within the memory range for the array // so it may be intentional. if (_settings->inconclusive) { arrayIndexOutOfBoundsError(tok, arrayInfo, indexes); break; // only warn about the first one } } } } } } // Loop.. else if (Token::simpleMatch(tok, "for (")) { bool bailout = false; checkScopeForBody(tok, arrayInfo, bailout); if (bailout) break; continue; } // Check function call.. if (Token::Match(tok, "%var% (")) { checkFunctionCall(tok, arrayInfo); } if (Token::Match(tok, "strncpy|memcpy|memmove ( %varid% , %str% , %num% )", arrayInfo.varid())) { unsigned int num = (unsigned int)MathLib::toLongNumber(tok->strAt(6)); if (Token::getStrLength(tok->tokAt(4)) >= (unsigned int)total_size && (unsigned int)total_size == num) { if (_settings->inconclusive) bufferNotZeroTerminatedError(tok, tok->strAt(2), tok->str()); } } if ((Token::Match(tok, "strncpy|strncat ( %varid% , %var% , %num% )", arrayInfo.varid())) || (Token::Match(tok, "strncpy|strncat ( %varid% , %var% [ %any% ] , %num% )", arrayInfo.varid()))) { const int offset = tok->strAt(5) == "[" ? 3 : 0; // check for strncpy which is not terminated if (tok->str() == "strncpy") { // strncpy takes entire variable length as input size unsigned int num = (unsigned int)MathLib::toLongNumber(tok->strAt(6 + offset)); if (num >= total_size) { const Token *tok2 = tok->next()->link()->next(); for (; tok2; tok2 = tok2->next()) { if (tok2->varId() == tok->tokAt(2)->varId()) { if (!Token::Match(tok2, "%varid% [ %any% ] = 0 ;", tok->tokAt(2)->varId())) { // this is currently 'inconclusive'. See TestBufferOverrun::terminateStrncpy3 if (_settings->isEnabled("style") && _settings->inconclusive) terminateStrncpyError(tok, tok->strAt(2)); } break; } } } } // Dangerous usage of strncat.. if (tok->str() == "strncat") { const MathLib::bigint n = MathLib::toLongNumber(tok->strAt(6 + offset)); if (n >= total_size) strncatUsageError(tok); } // Dangerous usage of strncpy + strncat.. if (Token::Match(tok->tokAt(8 + offset), "; strncat ( %varid% , %any% , %num% )", arrayInfo.varid())) { const MathLib::bigint n = MathLib::toLongNumber(tok->strAt(6 + offset)) + MathLib::toLongNumber(tok->strAt(15 + offset)); if (n > total_size) strncatUsageError(tok->tokAt(9 + offset)); } } // Writing data into array.. if (Token::Match(tok, "strcpy|strcat ( %varid% , %str% )", arrayInfo.varid())) { const std::size_t len = Token::getStrLength(tok->tokAt(4)); if (total_size > 0 && len >= (unsigned int)total_size) { bufferOverrunError(tok, arrayInfo.varname()); continue; } } // Detect few strcat() calls if (total_size > 0 && Token::Match(tok, "strcat ( %varid% , %str% ) ;", arrayInfo.varid())) { std::size_t charactersAppend = 0; const Token *tok2 = tok; while (tok2 && Token::Match(tok2, "strcat ( %varid% , %str% ) ;", arrayInfo.varid())) { charactersAppend += Token::getStrLength(tok2->tokAt(4)); if (charactersAppend >= (unsigned int)total_size) { bufferOverrunError(tok2, arrayInfo.varname()); break; } tok2 = tok2->tokAt(7); } } if (Token::Match(tok, "sprintf ( %varid% , %str% [,)]", arrayInfo.varid())) { checkSprintfCall(tok, total_size); } // snprintf.. if (total_size > 0 && Token::Match(tok, "snprintf ( %varid% , %num% ,", arrayInfo.varid())) { const MathLib::bigint n = MathLib::toLongNumber(tok->strAt(4)); if (n > total_size) outOfBoundsError(tok->tokAt(4), "snprintf size", true, n, total_size); } // readlink() / readlinkat() buffer usage if (_settings->standards.posix) { if (Token::Match(tok, "readlink ( %any% , %varid% , %num% )", arrayInfo.varid())) checkReadlinkBufferUsage(tok, scope_begin, total_size, false); else if (Token::Match(tok, "readlinkat ( %any% , %any% , %varid% , %num% )", arrayInfo.varid())) checkReadlinkBufferUsage(tok, scope_begin, total_size, true); } // undefined behaviour: result of pointer arithmetic is out of bounds if (_settings->isEnabled("portability") && Token::Match(tok, "= %varid% + %num% ;", arrayInfo.varid())) { const MathLib::bigint index = MathLib::toLongNumber(tok->strAt(3)); if (index < 0 || index > arrayInfo.num(0)) { pointerOutOfBoundsError(tok->next(), "array"); } } } } void CheckBufferOverrun::checkReadlinkBufferUsage(const Token* tok, const Token *scope_begin, const MathLib::bigint total_size, const bool is_readlinkat) { unsigned char param_offset = is_readlinkat ? 2 : 0; const std::string funcname = is_readlinkat ? "readlinkat" : "readlink"; const MathLib::bigint n = MathLib::toLongNumber(tok->strAt(6 + param_offset)); if (total_size > 0 && n > total_size) outOfBoundsError(tok->tokAt(4 + param_offset), funcname + "() buf size", true, n, total_size); if (!_settings->inconclusive) return; // readlink()/readlinkat() never terminates the buffer, check the end of the scope for buffer termination. bool found_termination = false; const Token *scope_end = scope_begin->link(); for (const Token *tok2 = tok->tokAt(8 + param_offset); tok2 && tok2 != scope_end; tok2 = tok2->next()) { if (Token::Match(tok2, "%varid% [ %any% ] = 0 ;", tok->tokAt(4 + param_offset)->varId())) { found_termination = true; break; } } if (!found_termination) { bufferNotZeroTerminatedError(tok, tok->strAt(4 + param_offset), funcname); } else if (n == total_size) { possibleReadlinkBufferOverrunError(tok, funcname, tok->strAt(4 + param_offset)); } } //--------------------------------------------------------------------------- // Checking local variables in a scope //--------------------------------------------------------------------------- void CheckBufferOverrun::checkGlobalAndLocalVariable() { // check string literals for (const Token *tok = _tokenizer->tokens(); tok; tok = tok->next()) { if (Token::Match(tok, "%str% [ %num% ]")) { std::string str = tok->strValue(); std::size_t index = std::atoi(tok->tokAt(2)->str().c_str()); if (index > str.length()) { bufferOverrunError(tok, tok->str()); } } } // check all known fixed size arrays first by just looking them up for (unsigned int i = 1; i <= _tokenizer->varIdCount(); i++) { const Variable *var = _tokenizer->getSymbolDatabase()->getVariableFromVarId(i); if (var && var->isArray() && var->dimension(0) > 0) { const ArrayInfo arrayInfo(var, _tokenizer); const Token *tok = var->nameToken(); while (tok && tok->str() != ";") { if (tok->str() == "{") { if (Token::simpleMatch(tok->previous(), "= {")) tok = tok->link(); else break; } tok = tok->next(); } if (!tok) break; if (tok->str() == "{") tok = tok->next(); checkScope(tok, arrayInfo); } } // find all dynamically allocated arrays next by parsing the token stream // Count { and } when parsing all tokens int indentlevel = 0; for (const Token *tok = _tokenizer->tokens(); tok; tok = tok->next()) { if (tok->str() == "{") ++indentlevel; else if (tok->str() == "}") --indentlevel; if (indentlevel <= 0) continue; // size : Max array index MathLib::bigint size = 0; // type : The type of a array element std::string type; // varid : The variable id for the array unsigned int varid = 0; // nextTok : number of tokens used in variable declaration - used to skip to next statement. int nextTok = 0; // if the previous token exists, it must be either a variable name or "[;{}]" if (tok->previous() && (!tok->previous()->isName() && !Token::Match(tok->previous(), "[;{}]"))) continue; _errorLogger->reportProgress(_tokenizer->getFiles()->front(), "Check (BufferOverrun::checkGlobalAndLocalVariable)", tok->progressValue()); if (Token::Match(tok, "[*;{}] %var% = new %type% [ %num% ]")) { size = MathLib::toLongNumber(tok->strAt(6)); type = tok->strAt(4); varid = tok->next()->varId(); nextTok = 8; } else if (Token::Match(tok, "[*;{}] %var% = new %type% ( %num% )")) { size = 1; type = tok->strAt(4); varid = tok->next()->varId(); nextTok = 8; } else if (Token::Match(tok, "[;{}] %var% = %str% ;") && tok->next()->varId() > 0 && NULL != Token::findmatch(_tokenizer->tokens(), "[;{}] const| %type% * %varid% ;", tok->next()->varId())) { size = 1 + int(tok->tokAt(3)->strValue().size()); type = "char"; varid = tok->next()->varId(); nextTok = 4; } else if (Token::Match(tok, "[*;{}] %var% = malloc|alloca ( %num% ) ;")) { size = MathLib::toLongNumber(tok->strAt(5)); type = "char"; // minimum type, typesize=1 varid = tok->next()->varId(); nextTok = 7; if (varid > 0) { // get type of variable const Variable *var = _tokenizer->getSymbolDatabase()->getVariableFromVarId(varid); /** @todo false negatives: this may be too conservative */ if (!var || var->typeEndToken()->str() != "*" || var->typeStartToken()->next() != var->typeEndToken()) continue; // get name of variable type = var->typeStartToken()->str(); // malloc() gets count of bytes and not count of // elements, so we should calculate count of elements // manually unsigned int sizeOfType = _tokenizer->sizeOfType(var->typeStartToken()); if (sizeOfType > 0) size /= static_cast(sizeOfType); } } else { continue; } if (varid == 0) continue; Token sizeTok(0); sizeTok.str(type); const MathLib::bigint total_size = size * static_cast(_tokenizer->sizeOfType(&sizeTok)); if (total_size == 0) continue; std::vector v; ArrayInfo temp(varid, tok->next()->str(), total_size / size, size); checkScope(tok->tokAt(nextTok), v, temp); } } //--------------------------------------------------------------------------- //--------------------------------------------------------------------------- // Checking member variables of structs.. //--------------------------------------------------------------------------- void CheckBufferOverrun::checkStructVariable() { const SymbolDatabase * symbolDatabase = _tokenizer->getSymbolDatabase(); std::list::const_iterator scope; // find every class and struct for (scope = symbolDatabase->scopeList.begin(); scope != symbolDatabase->scopeList.end(); ++scope) { // only check classes and structures if (!scope->isClassOrStruct()) continue; // check all variables to see if they are arrays std::list::const_iterator var; for (var = scope->varlist.begin(); var != scope->varlist.end(); ++var) { // find all array variables if (var->isArray()) { // create ArrayInfo from the array variable ArrayInfo arrayInfo(&*var, _tokenizer); std::list::const_iterator func_scope; // find every function for (func_scope = symbolDatabase->scopeList.begin(); func_scope != symbolDatabase->scopeList.end(); ++func_scope) { // only check functions if (func_scope->type != Scope::eFunction) continue; // check for member variables if (func_scope->functionOf == &*scope) { // only check non-empty function if (func_scope->function->start->next() != func_scope->function->start->link()) { // start checking after the { const Token *tok = func_scope->function->start->next(); checkScope(tok, arrayInfo); } } // skip inner scopes.. /** @todo false negatives: handle inner scopes someday */ if (scope->nestedIn->isClassOrStruct()) continue; std::vector varname; varname.push_back(""); varname.push_back(arrayInfo.varname()); // search the function and it's parameters for (const Token *tok3 = func_scope->classDef; tok3 && tok3 != func_scope->classEnd; tok3 = tok3->next()) { // search for the class/struct name if (tok3->str() != scope->className) continue; // Declare variable: Fred fred1; if (Token::Match(tok3->next(), "%var% ;")) varname[0] = tok3->strAt(1); // Declare pointer or reference: Fred *fred1 else if (Token::Match(tok3->next(), "*|& %var% [,);=]")) varname[0] = tok3->strAt(2); else continue; // check for variable sized structure if (scope->type == Scope::eStruct && var->isPublic()) { // last member of a struct with array size of 0 or 1 could be a variable sized structure if (var->dimensions().size() == 1 && var->dimension(0) < 2 && var->index() == (scope->varlist.size() - 1)) { // dynamically allocated so could be variable sized structure if (tok3->next()->str() == "*") { // check for allocation if ((Token::Match(tok3->tokAt(3), "; %var% = malloc ( %num% ) ;") || (Token::Match(tok3->tokAt(3), "; %var% = (") && Token::Match(tok3->linkAt(6), ") malloc ( %num% ) ;"))) && (tok3->strAt(4) == tok3->strAt(2))) { MathLib::bigint size; // find size of allocation if (tok3->strAt(3) == "(") // has cast size = MathLib::toLongNumber(tok3->linkAt(6)->strAt(3)); else size = MathLib::toLongNumber(tok3->strAt(8)); // We don't calculate the size of a structure even when we know // the size of the members. We just assign a length of 100 for // any struct. If the size is less than 100, we assume the // programmer knew the size and specified it rather than using // sizeof(struct). If the size is greater than 100, we assume // the programmer specified the size as sizeof(struct) + number. // Either way, this is just a guess and could be wrong. The // information to make the right decision has been simplified // away by the time we get here. if (size != 100) { // magic number for size of struct // check if a real size was specified and give up // malloc(10) rather than malloc(sizeof(struct)) if (size < 100) continue; // calculate real array size based on allocated size MathLib::bigint elements = (size - 100) / arrayInfo.element_size(); arrayInfo.num(0, arrayInfo.num(0) + elements); } } // size unknown so assume it is a variable sized structure else continue; } } } // Goto end of statement. const Token *CheckTok = NULL; while (tok3 && tok3 != func_scope->classEnd) { // End of statement. if (tok3->str() == ";") { CheckTok = tok3; break; } // End of function declaration.. if (Token::simpleMatch(tok3, ") ;")) break; // Function implementation.. if (Token::simpleMatch(tok3, ") {")) { CheckTok = tok3->tokAt(2); break; } tok3 = tok3->next(); } if (!tok3) break; if (!CheckTok) continue; // Check variable usage.. ArrayInfo temp = arrayInfo; temp.varid(0); // do variable lookup by variable and member names rather than varid std::string varnames; // use class and member name for messages for (unsigned int i = 0; i < varname.size(); ++i) varnames += (i == 0 ? "" : ".") + varname[i]; temp.varname(varnames); checkScope(CheckTok, varname, temp); } } } } } } //--------------------------------------------------------------------------- void CheckBufferOverrun::bufferOverrun() { checkGlobalAndLocalVariable(); checkStructVariable(); checkBufferAllocatedWithStrlen(); checkInsecureCmdLineArgs(); } //--------------------------------------------------------------------------- MathLib::bigint CheckBufferOverrun::countSprintfLength(const std::string &input_string, const std::list ¶meters) { bool percentCharFound = false; std::size_t input_string_size = 1; bool handleNextParameter = false; std::string digits_string = ""; bool i_d_x_f_found = false; std::list::const_iterator paramIter = parameters.begin(); std::size_t parameterLength = 0; for (std::string::size_type i = 0; i < input_string.length(); ++i) { if (input_string[i] == '\\') { if (input_string[i+1] == '0') break; ++input_string_size; ++i; continue; } if (percentCharFound) { switch (input_string[i]) { case 'f': case 'x': case 'X': case 'i': i_d_x_f_found = true; case 'c': case 'e': case 'E': case 'g': case 'o': case 'u': case 'p': case 'n': handleNextParameter = true; break; case 'd': i_d_x_f_found = true; if (paramIter != parameters.end() && *paramIter && (*paramIter)->str()[0] != '"') parameterLength = (*paramIter)->str().length(); handleNextParameter = true; break; case 's': if (paramIter != parameters.end() && *paramIter && (*paramIter)->str()[0] == '"') parameterLength = Token::getStrLength(*paramIter); handleNextParameter = true; break; } } if (input_string[i] == '%') percentCharFound = !percentCharFound; else if (percentCharFound) { digits_string.append(1, input_string[i]); } if (!percentCharFound) input_string_size++; if (handleNextParameter) { unsigned int tempDigits = static_cast(std::abs(std::atoi(digits_string.c_str()))); if (i_d_x_f_found) tempDigits = std::max(static_cast(tempDigits), 1U); if (digits_string.find('.') != std::string::npos) { const std::string endStr = digits_string.substr(digits_string.find('.') + 1); unsigned int maxLen = std::max(static_cast(std::abs(std::atoi(endStr.c_str()))), 1U); if (input_string[i] == 's') { // For strings, the length after the dot "%.2s" will limit // the length of the string. if (parameterLength > maxLen) parameterLength = maxLen; } else { // For integers, the length after the dot "%.2d" can // increase required length if (tempDigits < maxLen) tempDigits = maxLen; } } if (tempDigits < parameterLength) input_string_size += parameterLength; else input_string_size += tempDigits; parameterLength = 0; digits_string = ""; i_d_x_f_found = false; percentCharFound = false; handleNextParameter = false; if (paramIter != parameters.end()) ++paramIter; } } return (MathLib::bigint)input_string_size; } void CheckBufferOverrun::checkSprintfCall(const Token *tok, const MathLib::bigint size) { if (size == 0) return; std::list parameters; const Token* vaArg = tok->tokAt(2)->nextArgument()->nextArgument(); while (vaArg) { if (Token::Match(vaArg->next(), "[,)]")) { if (Token::Match(vaArg, "%str%")) parameters.push_back(vaArg); else if (Token::Match(vaArg, "%num%")) parameters.push_back(vaArg); else parameters.push_back(0); } else // Parameter is more complex than just a value or variable. Ignore it for now and skip to next token. parameters.push_back(0); vaArg = vaArg->nextArgument(); } MathLib::bigint len = countSprintfLength(tok->tokAt(2)->nextArgument()->strValue(), parameters); if (len > size) { bufferOverrunError(tok); } } //--------------------------------------------------------------------------- // Checking for allocating insufficient memory for copying a string by // allocating only strlen(src) bytes instead of strlen(src) + 1 bytes (one // extra for the terminating null character). // Example: // char *b = malloc(strlen(a)); // Should be malloc(strlen(a) + 1); // strcpy(b, a); // <== Buffer overrun //--------------------------------------------------------------------------- void CheckBufferOverrun::checkBufferAllocatedWithStrlen() { const char pattern[] = "%var% = new|malloc|g_malloc|g_try_malloc|realloc|g_realloc|g_try_realloc"; for (const Token *tok = Token::findmatch(_tokenizer->tokens(), pattern); tok; tok = Token::findmatch(tok->next(),pattern)) { unsigned int dstVarId; unsigned int srcVarId; // Look for allocation of a buffer based on the size of a string if (Token::Match(tok, "%var% = malloc|g_malloc|g_try_malloc ( strlen ( %var% ) )")) { dstVarId = tok->varId(); srcVarId = tok->tokAt(6)->varId(); tok = tok->tokAt(8); } else if (Token::Match(tok, "%var% = new char [ strlen ( %var% ) ]")) { dstVarId = tok->varId(); srcVarId = tok->tokAt(7)->varId(); tok = tok->tokAt(9); } else if (Token::Match(tok, "%var% = realloc|g_realloc|g_try_realloc ( %var% , strlen ( %var% ) )")) { dstVarId = tok->varId(); srcVarId = tok->tokAt(8)->varId(); tok = tok->tokAt(10); } else continue; // count { and } for tok int indentlevel = 0; for (; tok && tok->next(); tok = tok->next()) { // To avoid false positives and added complexity, we will only look for // improper usage of the buffer within the block that it was allocated if (tok->str() == "{") { ++indentlevel; } else if (tok->str() == "}") { --indentlevel; if (indentlevel < 0) return; } // If the buffers are modified, we can't be sure of their sizes if (tok->varId() == srcVarId || tok->varId() == dstVarId) break; if (Token::Match(tok, "strcpy ( %varid% , %var% )", dstVarId) && tok->tokAt(4)->varId() == srcVarId) { bufferOverrunError(tok); } else if (Token::Match(tok, "sprintf ( %varid% , %str% , %var% )", dstVarId) && tok->tokAt(6)->varId() == srcVarId && tok->strAt(4).find("%s") != std::string::npos) { bufferOverrunError(tok); } } } } //--------------------------------------------------------------------------- // Checking for buffer overflow caused by copying command line arguments // into fixed-sized buffers without checking to make sure that the command // line arguments will not overflow the buffer. // // int main(int argc, char* argv[]) // { // char prog[10]; // strcpy(prog, argv[0]); <-- Possible buffer overrun // } //--------------------------------------------------------------------------- void CheckBufferOverrun::checkInsecureCmdLineArgs() { const char pattern[] = "main ( int %var% , char *"; for (const Token *tok = Token::findmatch(_tokenizer->tokens(), pattern); tok; tok = Token::findmatch(tok->next(),pattern)) { // Get the name of the argv variable unsigned int varid = 0; if (Token::Match(tok, "main ( int %var% , char * %var% [ ] ,|)")) { varid = tok->tokAt(7)->varId(); } else if (Token::Match(tok, "main ( int %var% , char * * %var% ,|)")) { varid = tok->tokAt(8)->varId(); } if (varid == 0) continue; // Jump to the opening curly brace tok = tok->next()->link(); if (!Token::simpleMatch(tok, ") {")) continue; tok = tok->next(); // Search within main() for possible buffer overruns involving argv int indentlevel = -1; for (; tok; tok = tok->next()) { if (tok->str() == "{") { ++indentlevel; } else if (tok->str() == "}") { --indentlevel; if (indentlevel < 0) return; } // If argv is modified or tested, its size may be being limited properly if (tok->varId() == varid) break; // Match common patterns that can result in a buffer overrun // e.g. strcpy(buffer, argv[0]) if (Token::Match(tok, "strcpy|strcat ( %var% , * %varid%", varid) || Token::Match(tok, "strcpy|strcat ( %var% , %varid% [", varid)) { cmdLineArgsError(tok); } else if (Token::Match(tok, "sprintf ( %var% , %str% , %varid% [", varid) && tok->strAt(4).find("%s") != std::string::npos) { cmdLineArgsError(tok); } else if (Token::Match(tok, "sprintf ( %var% , %str% , * %varid%", varid) && tok->strAt(4).find("%s") != std::string::npos) { cmdLineArgsError(tok); } } } } //--------------------------------------------------------------------------- void CheckBufferOverrun::negativeIndexError(const Token *tok, MathLib::bigint index) { std::ostringstream ostr; ostr << "Array index " << index << " is out of bounds"; reportError(tok, Severity::error, "negativeIndex", ostr.str()); } void CheckBufferOverrun::negativeIndex() { const char pattern[] = "[ %num% ]"; for (const Token *tok = Token::findmatch(_tokenizer->tokens(), pattern); tok; tok = Token::findmatch(tok->next(),pattern)) { const MathLib::bigint index = MathLib::toLongNumber(tok->next()->str()); if (index < 0) { // Negative index. Check if it's an array. const Token *tok2 = tok; while (Token::simpleMatch(tok2->previous(), "]")) tok2 = tok2->previous()->link(); if (tok2->previous() && tok2->previous()->varId()) { const Variable *var = _tokenizer->getSymbolDatabase()->getVariableFromVarId(tok2->previous()->varId()); if (var && var->isArray()) negativeIndexError(tok, index); } } } } #include "executionpath.h" /// @addtogroup Checks /// @{ CheckBufferOverrun::ArrayInfo::ArrayInfo() { _element_size = 0; _varid = 0; } CheckBufferOverrun::ArrayInfo::ArrayInfo(const CheckBufferOverrun::ArrayInfo &ai) { *this = ai; } CheckBufferOverrun::ArrayInfo::ArrayInfo(const Variable *var, const Tokenizer *tokenizer) { _varid = var->varId(); _varname = var->name(); for (size_t i = 0; i < var->dimensions().size(); i++) _num.push_back(var->dimension(i)); if (var->typeEndToken()->str() == "*") _element_size = tokenizer->sizeOfType(var->typeEndToken()); else if (var->typeStartToken()->str() == "struct") _element_size = 100; else _element_size = tokenizer->sizeOfType(var->typeEndToken()); } CheckBufferOverrun::ArrayInfo & CheckBufferOverrun::ArrayInfo::operator=(const CheckBufferOverrun::ArrayInfo &ai) { if (&ai != this) { _element_size = ai._element_size; _num = ai._num; _varid = ai._varid; _varname = ai._varname; } return *this; } /** * Create array info with specified data * The intention is that this is only a temporary solution.. all * checking should be based on ArrayInfo from the start and then * this will not be needed as the declare can be used instead. */ CheckBufferOverrun::ArrayInfo::ArrayInfo(unsigned int id, const std::string &name, MathLib::bigint size1, MathLib::bigint n) { _element_size = size1; _num.push_back(n); _varid = id; _varname = name; } CheckBufferOverrun::ArrayInfo CheckBufferOverrun::ArrayInfo::limit(MathLib::bigint value) const { MathLib::bigint uvalue = std::max(MathLib::bigint(0), value); MathLib::bigint n = 1; for (unsigned int i = 0; i < _num.size(); ++i) n *= _num[i]; if (uvalue > n) n = uvalue; return ArrayInfo(_varid, _varname, _element_size, n - uvalue); } bool CheckBufferOverrun::ArrayInfo::declare(const Token *tok, const Tokenizer &tokenizer) { _num.clear(); _element_size = 0; _varname.clear(); if (!tok) return false; if (!tok->isName() || tok->str() == "return") return false; while (tok && (tok->str() == "static" || tok->str() == "const" || tok->str() == "extern")) tok = tok->next(); // ivar : number of type tokens int ivar = 0; if (Token::Match(tok, "%type% *| %var% [")) ivar = 1; else if (Token::Match(tok, "%type% %type% *| %var% [")) ivar = 2; else return false; if (tok->str().find(":") != std::string::npos) return false; // Goto variable name token, get element size.. const Token *vartok = tok->tokAt(ivar); if (vartok->str() == "*") { _element_size = tokenizer.sizeOfType(vartok); vartok = vartok->next(); } else if (tok->str() == "struct") { _element_size = 100; } else { _element_size = tokenizer.sizeOfType(tok); } _varname = vartok->str(); _varid = vartok->varId(); if (!_varid) return false; const Token *atok = vartok->tokAt(2); if (!Token::Match(atok, "%num% ] ;|=|[")) return false; while (Token::Match(atok, "%num% ] ;|=|[")) { _num.push_back(MathLib::toLongNumber(atok->str())); atok = atok->next(); if (Token::simpleMatch(atok, "] [")) atok = atok->tokAt(2); } if (Token::Match(atok, "] = !!{")) return false; return (!_num.empty() && Token::Match(atok, "] ;|=")); } /** * @brief %Check for buffer overruns (using ExecutionPath) */ class ExecutionPathBufferOverrun : public ExecutionPath { public: /** Startup constructor */ ExecutionPathBufferOverrun(Check *c, const std::map &arrayinfo) : ExecutionPath(c, 0), arrayInfo(arrayinfo), value(0) { } private: /** @brief Copy this check. Called from the ExecutionPath baseclass. */ ExecutionPath *copy() { return new ExecutionPathBufferOverrun(*this); } /** @brief is other execution path equal? */ bool is_equal(const ExecutionPath *e) const { const ExecutionPathBufferOverrun *c = static_cast(e); return (value == c->value); } /** @brief Buffer information */ const std::map &arrayInfo; /** no implementation => compiler error if used by accident */ void operator=(const ExecutionPathBufferOverrun &); /** internal constructor for creating extra checks */ ExecutionPathBufferOverrun(Check *c, const std::map &arrayinfo, unsigned int varid_) : ExecutionPath(c, varid_), arrayInfo(arrayinfo) { // Pretend that variables are initialized to 0 // This checking is not about uninitialized variables value = 0; } /** @brief Variable value. */ MathLib::bigint value; /** * @brief Assign value to a variable * @param checks the execution paths * @param varid the variable id * @param value the assigned value */ static void assign_value(std::list &checks, unsigned int varid, const std::string &value) { if (varid == 0) return; std::list::const_iterator it; for (it = checks.begin(); it != checks.end(); ++it) { ExecutionPathBufferOverrun *c = dynamic_cast(*it); if (c && c->varId == varid) c->value = MathLib::toLongNumber(value); } } /** * @brief Found array usage, analyse the array usage * @param tok token where usage occurs (only used when reporting the error) * @param checks The execution paths * @param varid1 variable id for the array * @param varid2 variable id for the index */ static void array_index(const Token *tok, std::list &checks, unsigned int varid1, unsigned int varid2) { if (checks.empty() || varid1 == 0 || varid2 == 0) return; // Locate array info corresponding to varid1 CheckBufferOverrun::ArrayInfo ai; { ExecutionPathBufferOverrun *c = dynamic_cast(checks.front()); std::map::const_iterator it; it = c->arrayInfo.find(varid1); if (it == c->arrayInfo.end()) return; ai = it->second; } // Check if varid2 variable has a value that is out-of-bounds std::list::const_iterator it; for (it = checks.begin(); it != checks.end(); ++it) { ExecutionPathBufferOverrun *c = dynamic_cast(*it); if (c && c->varId == varid2 && c->value >= ai.num(0)) { // variable value is out of bounds, report error CheckBufferOverrun *checkBufferOverrun = dynamic_cast(c->owner); if (checkBufferOverrun) { std::vector index; index.push_back(c->value); checkBufferOverrun->arrayIndexOutOfBoundsError(tok, ai, index); break; } } } } const Token *parse(const Token &tok, std::list &checks) const { if (Token::Match(tok.previous(), "[;{}]")) { // Declaring variable.. if (Token::Match(&tok, "%type% %var% ;") && tok.isStandardType()) { checks.push_back(new ExecutionPathBufferOverrun(owner, arrayInfo, tok.next()->varId())); return tok.tokAt(2); } // Assign variable.. if (Token::Match(&tok, "%var% = %num% ;")) { assign_value(checks, tok.varId(), tok.strAt(2)); return tok.tokAt(3); } } // Assign variable (unknown value = 0).. if (Token::Match(&tok, "%var% =")) { assign_value(checks, tok.varId(), "0"); return &tok; } // Assign variable (unknown value = 0).. if (Token::Match(tok.tokAt(-2), "(|, & %var% ,|)")) { assign_value(checks, tok.varId(), "0"); return &tok; } // Array index.. if (Token::Match(&tok, "%var% [ %var% ]")) { array_index(&tok, checks, tok.varId(), tok.tokAt(2)->varId()); return tok.tokAt(3); } return &tok; } }; /// @} void CheckBufferOverrun::executionPaths() { // Parse all variables and extract array info.. std::map arrayInfo; for (unsigned int i = 1; i <= _tokenizer->varIdCount(); i++) { const Variable *var = _tokenizer->getSymbolDatabase()->getVariableFromVarId(i); if (var && var->isArray() && var->dimension(0) > 0) arrayInfo[i] = ArrayInfo(var, _tokenizer); } // Perform checking - check how the arrayInfo arrays are used ExecutionPathBufferOverrun c(this, arrayInfo); checkExecutionPaths(_tokenizer->tokens(), &c); } void CheckBufferOverrun::arrayIndexThenCheck() { if (!_settings->isEnabled("style")) return; for (const Token *tok = _tokenizer->tokens(); tok; tok = tok->next()) { if (Token::Match(tok, "%var% [ %var% ]")) { const std::string arrayName(tok->str()); const std::string indexName(tok->strAt(2)); // skip array index.. tok = tok->tokAt(4); while (tok && tok->str() == "[") tok = tok->link()->next(); // syntax error if (!tok) return; // skip comparison if (Token::Match(tok, "==|!=|<|<=|>|>= %any% &&")) tok = tok->tokAt(2); // check if array index is ok if (Token::Match(tok, ("&& " + indexName + " <|<=").c_str())) arrayIndexThenCheckError(tok, indexName); } } } void CheckBufferOverrun::arrayIndexThenCheckError(const Token *tok, const std::string &indexName) { reportError(tok, Severity::style, "arrayIndexThenCheck", "Array index " + indexName + " is used before limits check\n" "Defensive programming: The variable " + indexName + " is used as array index and then there is a check that it is within limits. This can " "mean that the array might be accessed out-of-bounds. Reorder conditions such as '(a[i] && i < 10)' to '(i < 10 && a[i])'. That way the " "array will not be accessed when the index is out of limits."); }