DCL34-C. Use volatile for data that cannot be cached

An object that has volatile-qualified type may be modified in ways unknown to the implementation or have other unknown side effects. Asynchronous signal handling, for example, may cause objects to be modified in a manner unknown to the compiler. Without this type qualifier, unintended optimizations may occur. These optimizations may cause race conditions because a programmer may write code that prevents a race condition, yet the compiler is not aware of the programmer's data model and may modify the code during compilation to permit race conditions.

The volatile keyword eliminates this confusion by imposing restrictions on access and caching. According to the C99 Rationale [ISO/IEC 2003],

No caching through this lvalue: each operation in the abstract semantics must be performed (that is, no caching assumptions may be made, because the location is not guaranteed to contain any previous value). In the absence of this qualifier, the contents of the designated location may be assumed to be unchanged except for possible aliasing.

Type qualifying objects as volatile does not guarantee synchronization between multiple threads, protect against simultaneous memory accesses, or, unless used to declare objects of type sig_atomic_t, guarantee atomicity of accesses to the object. For restrictions specific to signal handlers, see SIG31-C. Do not access shared objects in signal handlers.

Noncompliant Code Example (Missing volatile)

This noncompliant code example relies on the reception of a SIGINT signal to toggle a flag to terminate a loop. However, because interrupted is not declared volatile, the read from it in main() may be optimized away by the compiler despite the assignment to the variable in the signal handler, and the loop may never terminate. When compiled on GCC with the -O optimization flag, for example, the program fails to terminate even upon receiving a SIGINT.

#include <signal.h>

sig_atomic_t interrupted;   /* Bug: not declared volatile */

void sigint_handler(int signum) {
  interrupted = 1;   /* Assignment may not be visible in main() */
}

int main(void) {
  signal(SIGINT, sigint_handler);
  while (!interrupted) {   /* Loop may never terminate */
   /* Do something */
  }
  return 0;
}

Compliant Solution

By adding the volatile qualifier to the variable declaration, interrupted is guaranteed to be accessed from its original address for every iteration of the while loop as well as from within the signal handler.

#include <signal.h>

volatile sig_atomic_t interrupted;

void sigint_handler(int signum) {
  interrupted = 1;
}

int main(void) {
  signal(SIGINT, sigint_handler);
  while (!interrupted) {
   /* Do something */
  }
  return 0;
}

The sig_atomic_t type is the integer type of an object that can be accessed as an atomic entity even in the presence of asynchronous interrupts. The type of sig_atomic_t is implementation-defined, though it provides some guarantees. Integer values ranging from SIG_ATOMIC_MIN through SIG_ATOMIC_MAX may be safely stored to a variable of the type. In addition, when sig_atomic_t is a signed integer type, SIG_ATOMIC_MIN must be no greater than -127 and SIG_ATOMIC_MAX no less than 127. Otherwise, SIG_ATOMIC_MIN must be 0 and SIG_ATOMIC_MAX must be no less than 255. The macros SIG_ATOMIC_MIN and SIG_ATOMIC_MAX are defined in the header <stdint.h>.

Noncompliant Code Example (Cast to volatile)

In this noncompliant code example, the thread_func function runs in the context of multiple threads designed to communicate with one another via the global variable end_processing. The function attempts to prevent the compiler from optimizing away the while loop condition by casting the variable to volatile before accessing it. However, because end_processing is not declared volatile, the assignment to it in the body of the loop does not need to be flushed from a register into memory and consequently may not be visible when read despite the cast. As a result, the loop may never terminate.

extern int compute(void*);

static _Bool end_processing;

int thread_func(void *arg) {
  while (0 == *(volatile _Bool*)&end_processing) {
    int status;
    status = compute(arg);
    if (status) {
      /* Notify other threads to end processing */
      end_processing = 1;
      break;
    }
  }
  return 0;
}

Compliant Solution

By adding the volatile qualifier to the variable declaration, end_processing is guaranteed to be read from and written to its original address for every iteration of the while loop.

extern int compute(void*);

static volatile _Bool end_processing;

int thread_func(void *arg) {
  while (0 == end_processing) {
    int status;
    status = compute(arg);
    if (status) {
      /* Notify other threads to end processing */
      end_processing = 1;
      break;
    }
  }
  return 0;
}

Note, however, that declaring an object volatile is not sufficient to prevent data races when the object is simultaneously accessed from within two or more threads of execution. Additional memory visibility constraints may necessitate the use of platform-specific constructs such as memory barriers, for example, when each of the threads runs on a different processor. See CON02-C. Do not use volatile as a synchronization primitive for more information.

Risk Assessment

Failing to use the volatile qualifier can result in race conditions in asynchronous portions of the code, causing unexpected values to be stored and leading to possible data integrity violations.

Automated Detection

Related Vulnerabilities

Search for vulnerabilities resulting from the violation of this rule on the CERT website.

Related Guidelines

Bibliography