Page History

If a lock is being held and an operation that can block is performed, any other thread that needs to acquire that lock may also block. This condition can degrade system performance or cause a deadlock to occur.  Blocking calls include, but are not limited to, network, file, and console I/O.

Using a blocking operation while holding a lock may be unavoidable for a portable solution. For instance, file access could be protected via a lock to prevent multiple threads from mutating the contents of the file. Or, a thread may be required to block while holding one or more locks and waiting to acquire another lock. In these cases, attempt to hold the lock for the least time required. Additionally, if acquiring multiple locks, the order of locking must avoid deadlock, as specified in CON35-C. Avoid deadlock by locking in a predefined order.

Noncompliant Code Example

This noncompliant code example demonstrates an occurrence of a blocking call that waits to receive data on a socket example calls fopen() while a mutex is locked. The recv() call blocks until data arrives on the socket. While it calls to fopen() and fclose() are blocking and may block for an extended period of time if the file resides on a network drive. While the call is blocked, other threads that are waiting for the lock are also blocked.Although this example is specific to network I/O, the recv() call could be replaced with any blocking call, and the same behavior would occur.

#include <stdio.h>
#include <threads.h>
 
mtx_t mutex;

voidint thread_foo(void *ptr) {
  uint32_t num;
  int result;
  intFILE sock*fp;

  /* sock is a connected TCP socket */
  if ((result = mtx_lock(&mutex)) != thrd_success) {
    /* Handle Errorerror */
  }
 
  iffp ((result = recvfopen(sock, (void *)&num, sizeof(uint32_t), 0)) < 0"SomeNetworkFile", "r");
  if (fp != NULL) {
    /* Handle ErrorWork with the file */
  }

  /* ... */
fclose(fp);
  }
 
  if ((result = mtx_unlock(&mutex)) != thrd_success) {
    /* Handle Errorerror */
  }

  return 0;
}

int main(void) {
  thrd_t thread;
  int result;

  if ((result = mtx_init(&mutex, mtx_plain)) != thrd_success) {
    /* Handle Errorerror */
  }

  if (thrd_create(&thread,(void *)& thread_foo, NULL) != thrd_success) {
    /* Handle Errorerror */
  }

  /* ... */

  if (thrd_join(thread, NULL);

  if ((result = mtx_destroy(&mutex)) != thrd_success) {
    /* Handle Errorerror */
  }

  mtx_destroy(&mutex);

  return 0;
}

Compliant Solution (Block while Not Locked)

This compliant solution performs the recv() call file operations when the lock has not been acquired. The blocking behavior consequently affects only the thread that called the blocking function.

void thread_foo(void *ptr) {
  uint32#include <stdio.h>
#include <threads.h>
 
mtx_t nummutex;
  int result;
  int sock;

  /* sock is a connected TCP socket */

  if ((result = recv(sock, (void *)&num, sizeof(uint32_t), 0)) < 0)  
int thread_foo(void *ptr) {
  int result;
 /* HandleFILE Error */
  }

  if ((result = mtx_lock(&mutex)) != thrd_success) {
    /* Handle Error */
  }

  /* ... */
fp = fopen("SomeNetworkFile", "r");
 
  if ((result = pthread_mutex_unlock(&mutex))fp != 0NULL) {
    /* Handle Error */
  }
}

Compliant Solution (Use a Nonblocking Call)

This compliant solution performs the recv() call with the parameter o_nonblock, which causes the call to fail if no messages are available on the socket.

void thread_foo(void *ptr) {
  uint32_t num;
  int result;

  /* sock is a connected TCP socket */

  if ((result = recv(sock, (void *)&num, sizeof(uint32_t), O_NONBLOCK)) < 0) {
    /* Handle Error */
  Work with the file */
    fclose(fp);
  }

  if ((result = mtx_lock(&mutex)) != thrd_success) {
    /* Handle Errorerror */
  }

  /* ... */

  if ((result = mtxpthread_mutex_unlock(&mutex)) != thrd_success0) {
    /* Handle Errorerror */
  }

  return 0;
}

...

CON36-EX1: A thread may block while holding one or more locks and waiting to acquire another lock. When acquiring multiple locks, the order of locking must avoid deadlock, as specified in CON35-C. Avoid deadlock by locking in predefined order.

Risk Assessment

Blocking or lengthy operations performed within synchronized regions could result in a deadlocked or an unresponsive system.

Related Guidelines

Automated Detection

Related Vulnerabilities

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

Related Guidelines

Key here (explains table format and definitions)

...

...

...

Sources

...

...

Image Modified Image Modified Image Modified