1# Commonly recurring errors in bmcweb 2 3What follows is a list of common errors that new users to bmcweb tend to make 4when operating within its bounds for the first time. If this is your first time 5developing in bmcweb, the maintainers highly recommend reading and understanding 6_all_ of common traps before continuing with any development. Every single one 7of the examples below compile without warnings, but are incorrect in 8not-always-obvious ways, or impose a pattern that tends to cause hard to find 9bugs, or bugs that appear later. Every one has been submitted to code review 10multiple times. 11 12### 1. Directly dereferencing a pointer without checking for validity first 13```C++ 14int myBadMethod(const nlohmann::json& j){ 15 const int* myPtr = j.get_if<int>(); 16 return *myPtr; 17} 18``` 19This pointer is not guaranteed to be filled, and could be a null dereference. 20 21### 2. String views aren't null terminated 22```C++ 23int getIntFromString(const std::string_view s){ 24 return std::atoi(s.data()); 25} 26``` 27This will give the right answer much of the time, but has the possibility to 28fail when string\_view is not null terminated. Use from\_chars instead, which 29takes both a pointer and a length 30 31### 3. Not handling input errors 32```C++ 33int getIntFromString(const std::string& s){ 34 return std::atoi(s.c_str()); 35} 36``` 37In the case where the string is not representable as an int, this will trigger 38undefined behavior at system level. Code needs to check for validity of the 39string, ideally with something like from\_chars, and return the appropriate error 40code. 41 42### 4. Walking off the end of a string 43```C++ 44std::string getFilenameFromPath(const std::string& path){ 45 size_t index = path.find("/"); 46 if (index != std::string::npos){ 47 // If the string ends with "/", this will walk off the end of the string. 48 return path.substr(pos + 1); 49 } 50 return ""; 51} 52``` 53 54### 5. Using methods that throw (or not handling bad inputs) 55```C++ 56int myBadMethod(nlohmann::json& j){ 57 return j.get<int>(); 58} 59``` 60This method throws, and bad inputs will not be handled 61 62Commonly used methods that fall into this pattern: 63 64- std::variant::get 65- std::vector::at 66- std::map::at 67- std::set::at 68- std::\<generic container type\>::at 69- nlohmann::json::operator!= 70- nlohmann::json::operator+= 71- nlohmann::json::at 72- nlohmann::json::get 73- nlohmann::json::get\_ref 74- nlohmann::json::get\_to 75- nlohmann::json::operator\<\< 76- nlohmann::json::operator\>\> 77- std::filesystem::create\_directory 78- std::filesystem::rename 79- std::filesystem::file\_size 80- std::stoi 81- std::stol 82- std::stoll 83 84#### Special note: JSON 85 86`nlohmann::json::parse` by default 87[throws](https://json.nlohmann.me/api/basic_json/parse/) on failure, but 88also accepts an optional argument that causes it to not throw: set the 3rd 89argument to `false`. 90 91`nlohmann::json::dump` by default 92[throws](https://json.nlohmann.me/api/basic_json/dump/) on failure, but also 93accepts an optional argument that causes it to not throw: set the 4th 94argument to `replace`. Although `ignore` preserves content 1:1, `replace` 95is preferred from a security point of view. 96 97#### Special note: Boost 98there is a whole class of boost asio functions that provide both a method that 99throws on failure, and a method that accepts and returns an error code. This is 100not a complete list, but users should verify in the boost docs when calling into 101asio methods, and prefer the one that returns an error code instead of throwing. 102 103- boost::asio::ip::tcp::acceptor::bind(); 104- boost::asio::ip::tcp::acceptor::cancel(); 105- boost::asio::ip::tcp::acceptor::close(); 106- boost::asio::ip::tcp::acceptor::listen(); 107- boost::asio::ip::address::make\_address(); 108 109### 6. Blocking functions 110 111bmcweb uses a single reactor for all operations. Blocking that reactor for any 112amount of time causes all other operations to stop. The common blocking 113functions that tend to be called incorrectly are: 114 115- sleep() 116- boost::asio::ip::tcp::socket::read() 117- boost::asio::ip::tcp::socket::read\_some() 118- boost::asio::ip::tcp::socket::write() 119- boost::asio::ip::tcp::socket::write\_some() 120- boost::asio::ip::tcp::socket::connect() 121- boost::asio::ip::tcp::socket::send() 122- boost::asio::ip::tcp::socket::wait() 123- boost::asio::steady\_timer::wait() 124 125Note: an exception is made for filesystem/disk IO read and write. This is 126mostly due to not having great abstractions for it that mate well with the async 127system, the fact that most filesystem accesses are into tmpfs (and therefore 128should be "fast" most of the time) and in general how little the filesystem is 129used in practice. 130 131### 7. Lack of locking between subsequent calls 132While global data structures are discouraged, they are sometimes required to 133store temporary state for operations that require it. Given the single 134threaded nature of bmcweb, they are not required to be explicitly threadsafe, 135but they must be always left in a valid state, and checked for other uses 136before occupying. 137 138```C++ 139std::optional<std::string> currentOperation; 140void firstCallbackInFlow(){ 141 currentOperation = "Foo"; 142} 143void secondCallbackInFlow(){ 144 currentOperation.reset(); 145} 146``` 147 148In the above case, the first callback needs a check to ensure that 149currentOperation is not already being used. 150 151### 8. Wildcard reference captures in lambdas 152``` 153std::string x; auto mylambda = [&](){ 154 x = "foo"; 155} 156do_async_read(mylambda) 157``` 158 159Numerous times, lifetime issues of const references have been injected into 160async bmcweb code. While capturing by reference can be useful, given how 161difficult these types of bugs are to triage, bmcweb explicitly requires that all 162code captures variables by name explicitly, and calls out each variable being 163captured by value or by reference. The above prototypes would change to 164[&x]()... Which makes clear that x is captured, and its lifetime needs tracked. 165 166 167### 9. URLs should end in "/" 168```C++ 169BMCWEB("/foo/bar"); 170``` 171Unless you explicitly have a reason not to (as there is one known exception 172where the behavior must differ) all URL handlers should end in "/". The bmcweb 173route handler will detect routes ending in slash and generate routes for both 174the route ending in slash and the one without. This allows both URLs to be 175used by users. While many specifications do not require this, it resolves a 176whole class of bug that we've seen in the past. 177 178 179### 10. URLs constructed in aggregate 180```C++ 181std::string routeStart = "/redfish/v1"; 182 183BMCWEB_ROUTE(routestart + "/SessionService/") 184``` 185Very commonly, bmcweb maintainers and contributors alike have to do audits of 186all routes that are available, to verify things like security and documentation 187accuracy. While these processes are largely manual, they can mostly be 188conducted by a simple grep statement to search for urls in question. Doing the 189above makes the route handlers no longer greppable, and complicates bmcweb 190patchsets as a whole. 191 192### 11. Not responding to 404 193```C++ 194BMCWEB_ROUTE("/myendpoint/<str>", 195 [](Request& req, Response& res, const std::string& id){ 196 crow::connections::systemBus->async_method_call( 197 [asyncResp](const boost::system::error_code ec, 198 const std::string& myProperty) { 199 if (ec) 200 { 201 messages::internalError(asyncResp->res); 202 return; 203 } 204 ... handle code 205 }, 206 "xyz.openbmc_project.Logging", 207 "/xyz/openbmc_project/mypath/" + id, 208 "xyz.MyInterface", "GetAll", ""); 209}); 210``` 211All bmcweb routes should handle 404 (not found) properly, and return it where 212appropriate. 500 internal error is not a substitute for this, and should be 213only used if there isn't a more appropriate error code that can be returned. 214This is important, because a number of vulnerability scanners attempt injection 215attacks in the form of /myendpoint/foobar, or /myendpoint/#$*(%)&#%$)(*& in an 216attempt to circumvent security. If the server returns 500 to any of these 217requests, the security scanner logs it as an error for followup. While in 218general these errors are benign, and not actually a real security threat, having 219a clean security run allows maintainers to minimize the amount of time spent 220triaging issues reported from these scanning tools. 221 222An implementation of the above that handles 404 would look like: 223```C++ 224BMCWEB_ROUTE("/myendpoint/<str>", 225 [](Request& req, Response& res, const std::string& id){ 226 crow::connections::systemBus->async_method_call( 227 [asyncResp](const boost::system::error_code ec, 228 const std::string& myProperty) { 229 if (ec == <error code that gets returned by not found>){ 230 messages::resourceNotFound(res); 231 return; 232 } 233 if (ec) 234 { 235 messages::internalError(asyncResp->res); 236 return; 237 } 238 ... handle code 239 }, 240 "xyz.openbmc_project.Logging", 241 "/xyz/openbmc_project/mypath/" + id, 242 "xyz.MyInterface", "GetAll", ""); 243}); 244``` 245 246Note: A more general form of this rule is that no handler should ever return 500 247on a working system, and any cases where 500 is found, can immediately be 248assumed to be [a bug in either the system, or bmcweb.](https://github.com/openbmc/bmcweb/blob/master/DEVELOPING.md#error-handling) 249 250### 12. Imprecise matching 251```C++ 252void isInventoryPath(const std::string& path){ 253 if (path.find("inventory")){ 254 return true; 255 } 256 return false; 257} 258``` 259When matching dbus paths, HTTP fields, interface names, care should be taken to 260avoid doing direct string containment matching. Doing so can lead to errors 261where fan1 and fan11 both report to the same object, and cause behavior breaks 262in subtle ways. 263 264When using dbus paths, rely on the methods on sdbusplus::message::object\_path. 265When parsing HTTP field and lists, use the RFC7230 implementations from 266boost::beast. 267 268Other commonly misused methods are: 269boost::iequals. Unless the standard you're implementing (as is the case in some 270HTTP fields) requires case insensitive comparisons, casing should be obeyed, 271especially when relying on user-driven data. 272 273- boost::starts\_with 274- boost::ends\_with 275- std::string::starts\_with 276- std::string::ends\_with 277- std::string::rfind 278 279The above methods tend to be misused to accept user data and parse various 280fields from it. In practice, there tends to be better, purpose built methods 281for removing just the field you need. 282 283### 13. Complete replacement of the response object 284 285``` 286void getMembers(crow::Response& res){ 287 res.jsonValue = {{"Value", 2}}; 288} 289``` 290 291In many cases, bmcweb is doing multiple async actions in parallel, and 292orthogonal keys within the Response object might be filled in from another 293task. Completely replacing the json object can lead to convoluted situations 294where the output of the response is dependent on the _order_ of the asynchronous 295actions completing, which cannot be guaranteed, and has many time caused bugs. 296 297``` 298void getMembers(crow::Response& res){ 299 res.jsonValue["Value"] = 2; 300} 301``` 302 303As an added benefit, this code is also more efficient, as it avoids the 304intermediate object construction and the move, and as a result, produces smaller 305binaries. 306 307Note, another form of this error involves calling nlohmann::json::reset(), to 308clear an object that's already been filled in. This has the potential to clear 309correct data that was already filled in from other sources. 310 311