The C++14 Standard, [support.runtime], paragraph 10 [ISO/IEC 14882-2014], states the following:
The common subset of the C and C++ languages consists of all declarations, definitions, and expressions that may appear in a well-formed C++ program and also in a conforming C program. A POF (“plain old function”) is a function that uses only features from this common subset, and that does not directly or indirectly use any function that is not a POF, except that it may use plain lock-free atomic operations. A plain lock-free atomic operation is an invocation of a function f from Clause 29, such that f is not a member function, and either f is the function atomic_is_lock_free, or for every atomic argument A passed to f, atomic_is_lock_free(A) yields true. All signal handlers shall have C linkage. The behavior of any function other than a POF used as a signal handler in a C++ program is implementation-defined.228
Footnote 228 states the following:
In particular, a signal handler using exception handling is very likely to have problems. Also, invoking
std::exit
may cause destruction of objects, including those of the standard library implementation, which, in general, yields undefined behavior in a signal handler.
If your signal handler is not a plain old function, then the behavior of a call to it in response to a signal is implementation-defined, at best, and is likely to result in undefined behavior. All signal handlers must meet the definition of a plain old function. In addition to the restrictions placed on signal handlers in a C program, this definition also prohibits the use of features that exist in C++ but not in C (such as non-POD [non–plain old data] objects and exceptions). This includes indirect use of such features through function calls.
In C++17, the wording has changed and relaxed some of the constraints on signal handlers. Section [support.signal], paragraph 3 says:
An evaluation is signal-safe unless it includes one of the following:
— a call to any standard library function, except for plain lock-free atomic operations and functions explicitly identified as signal-safe. [ Note: This implicitly excludes the use of new and delete expressions that rely on a library-provided memory allocator. — end note ]
— an access to an object with thread storage duration;
— a dynamic_cast expression;
— throwing of an exception;
— control entering a try-block or function-try-block;
— initialization of a variable with static storage duration requiring dynamic initialization (6.6.3, 9.7)220; or
— waiting for the completion of the initialization of a variable with static storage duration (9.7).A signal handler invocation has undefined behavior if it includes an evaluation that is not signal-safe.
Signal handlers in code that will be executed on C++17-compliant platforms must be signal-safe.
Noncompliant Code Example
In this noncompliant code example, the signal handler is declared as a static
function. However, since all signal handler functions must have C language linkage, and C++ is the default language linkage for functions in C++, calling the signal handler results in undefined behavior.
#include <csignal> static void sig_handler(int sig) { // Implementation details elided. } void install_signal_handler() { if (SIG_ERR == std::signal(SIGTERM, sig_handler)) { // Handle error } }
Compliant Solution
This compliant solution defines sig_handler()
as having C language linkage. As a consequence of declaring the signal handler with C language linkage, the signal handler will have external linkage rather than internal linkage.
#include <csignal> extern "C" void sig_handler(int sig) { // Implementation details elided. } void install_signal_handler() { if (SIG_ERR == std::signal(SIGTERM, sig_handler)) { // Handle error } }
Noncompliant Code Example
In this noncompliant code example, a signal handler calls a function that allows exceptions, and it attempts to handle any exceptions thrown. Because exceptions are not part of the common subset of C and C++ features, this example results in implementation-defined behavior. However, it is unlikely that the implementation's behavior will be suitable. For instance, on a stack-based architecture where a signal is generated asynchronously (instead of as a result of a call to std:abort()
or std::raise()
), it is possible that the stack frame is not properly initialized, causing stack tracing to be unreliable and preventing the exception from being caught properly.
#include <csignal> static void g() noexcept(false); extern "C" void sig_handler(int sig) { try { g(); } catch (...) { // Handle error } } void install_signal_handler() { if (SIG_ERR == std::signal(SIGTERM, sig_handler)) { // Handle error } }
Compliant Solution
There is no compliant solution whereby g()
can be called from the signal handler because it allows exceptions. Even if g()
were implemented such that it handled all exceptions and was marked noexcept(true)
, it would still be noncompliant to call g()
from a signal handler because g()
would still use a feature that is not a part of the common subset of C and C++ features allowed by a signal handler. Therefore, this compliant solution removes the call to g()
from the signal handler and instead polls a variable of type volatile sig_atomic_t
periodically; if the variable is set to 1
in the signal handler, then g()
is called to respond to the signal.
#include <csignal> volatile sig_atomic_t signal_flag = 0; static void g() noexcept(false); extern "C" void sig_handler(int sig) { signal_flag = 1; } void install_signal_handler() { if (SIG_ERR == std::signal(SIGTERM, sig_handler)) { // Handle error } } // Called periodically to poll the signal flag. void poll_signal_flag() { if (signal_flag == 1) { signal_flag = 0; try { g(); } catch(...) { // Handle error } } }
Risk Assessment
Failing to use a plain old function as a signal handler can result in implementation-defined behavior as well as undefined behavior. Given the number of features that exist in C++ that do not also exist in C, the consequences that arise from failure to comply with this rule can range from benign (harmless) behavior to abnormal program termination, or even arbitrary code execution.
Rule | Severity | Likelihood | Remediation Cost | Priority | Level |
---|---|---|---|---|---|
MSC54-CPP | High | Probable | High | P6 | L2 |
Automated Detection
Tool | Version | Checker | Description |
---|---|---|---|
Helix QAC | 2024.3 | C++2888 | |
Klocwork | 2024.3 | CERT.MSC.SIG_HANDLER.POF | |
Parasoft C/C++test | 2023.1 | CERT_CPP-MSC54-a | Properly define signal handlers |
Polyspace Bug Finder | R2024a | Checks for unsafe signal handlers (rule fully covered) |
Related Vulnerabilities
Search for vulnerabilities resulting from the violation of this rule on the CERT website.
Related Guidelines
SEI CERT C Coding Standard | SIG30-C. Call only asynchronous-safe functions within signal handlers SIG31-C. Do not access shared objects in signal handlers |
Bibliography
[ISO/IEC 14882-2014] | Subclause 18.10, "Other Runtime Support" |
3 Comments
Aaron Ballman
C++17 relaxes a lot of the rules regarding what is signal safe and not, so this rule is going to be a bit more strict than it needs to be once C++17 is ratified this year. For instance, it no longer talks about "plain old function", but is instead based on signal-safe executions.
Aaron Ballman
FYI, these are the changes to C++ that impact this rule: http://wg21.link/p0270 and the current wording is causing some consternation in a review for a checker that would cover the rule: https://reviews.llvm.org/D33825
David Svoboda
The C++14 material is still good, as we still plan to support it. But I've added text that should apply to C++17 and newer. thanks for the notification!