[c] How do I change a TCP socket to be non-blocking?

How do you make a socket non-blocking?

I am aware of the fcntl() function, but I've heard it's not always reliable.

This question is related to c sockets

The answer is


If you want to change socket to non blocking , precisely accept() to NON-Blocking state then

    int flags=fcntl(master_socket, F_GETFL);
        fcntl(master_socket, F_SETFL,flags| O_NONBLOCK); /* Change the socket into non-blocking state  F_SETFL is a command saying set flag and flag is 0_NONBLOCK     */                                          

while(1){
    if((newSocket = accept(master_socket, (struct sockaddr *) &address, &addr_size))<0){
                if(errno==EWOULDBLOCK){
                       puts("\n No clients currently available............ \n");
                       continue;
               }
    }else{
          puts("\nClient approched............ \n");
    }

}

I know it's an old question, but for everyone on google ending up here looking for information on how to deal with blocking and non-blocking sockets here is an in depth explanation of the different ways how to deal with the I/O modes of sockets - http://dwise1.net/pgm/sockets/blocking.html.

Quick summary:

  • So Why do Sockets Block?

  • What are the Basic Programming Techniques for Dealing with Blocking Sockets?

    • Have a design that doesn't care about blocking
    • Using select
    • Using non-blocking sockets.
    • Using multithreading or multitasking

fcntl() has always worked reliably for me. In any case, here is the function I use to enable/disable blocking on a socket:

#include <fcntl.h>

/** Returns true on success, or false if there was an error */
bool SetSocketBlockingEnabled(int fd, bool blocking)
{
   if (fd < 0) return false;

#ifdef _WIN32
   unsigned long mode = blocking ? 0 : 1;
   return (ioctlsocket(fd, FIONBIO, &mode) == 0) ? true : false;
#else
   int flags = fcntl(fd, F_GETFL, 0);
   if (flags == -1) return false;
   flags = blocking ? (flags & ~O_NONBLOCK) : (flags | O_NONBLOCK);
   return (fcntl(fd, F_SETFL, flags) == 0) ? true : false;
#endif
}

On Linux and BSD you can directly create the socket in non-blocking mode (https://www.man7.org/linux/man-pages/man7/socket.7.html):

int fd = socket(res->ai_family, res->ai_socktype | SOCK_NONBLOCK | SOCK_CLOEXEC, res->ai_protocol);
if (fd == -1) {
    perror("socket");
    return -1;
}

When accepting connections you can use the accept4 function to directly accept new connections in non-blocking mode (https://man7.org/linux/man-pages/man2/accept.2.html):

int fd = accept4(lfd, NULL, 0, SOCK_NONBLOCK | SOCK_CLOEXEC);
if (fd == -1) {
    perror("accept4");
    return -1;
}

I don't know why the accepted answer doesn't mention this.


The best method for setting a socket as non-blocking in C is to use ioctl. An example where an accepted socket is set to non-blocking is following:

long on = 1L;
unsigned int len;
struct sockaddr_storage remoteAddress;
len = sizeof(remoteAddress);
int socket = accept(listenSocket, (struct sockaddr *)&remoteAddress, &len)
if (ioctl(socket, (int)FIONBIO, (char *)&on))
{
    printf("ioctl FIONBIO call failed\n");
}

Generally you can achieve the same effect by using normal blocking IO and multiplexing several IO operations using select(2), poll(2) or some other system calls available on your system.

See The C10K problem for the comparison of approaches to scalable IO multiplexing.


It is sometimes convenient to employ the "send/recv" family of system calls. If the flags parameter contains the MSG_DONTWAIT flag, each call will behave similar to a socket having the O_NONBLOCK flag set.

ssize_t send(int sockfd, const void *buf, size_t len, int flags);

ssize_t recv(int sockfd, void *buf, size_t len, int flags);

fcntl() or ioctl() are used to set the properties for file streams. When you use this function to make a socket non-blocking, function like accept(), recv() and etc, which are blocking in nature will return error and errno would be set to EWOULDBLOCK. You can poll file descriptor sets to poll on sockets.


You're misinformed about fcntl() not always being reliable. It's untrue.

To mark a socket as non-blocking the code is as simple as:

// where socketfd is the socket you want to make non-blocking
int status = fcntl(socketfd, F_SETFL, fcntl(socketfd, F_GETFL, 0) | O_NONBLOCK);

if (status == -1){
  perror("calling fcntl");
  // handle the error.  By the way, I've never seen fcntl fail in this way
}

Under Linux, on kernels > 2.6.27 you can also create sockets non-blocking from the outset using socket() and accept4().

e.g.

   // client side
   int socketfd = socket(AF_INET, SOCK_STREAM | SOCK_NONBLOCK, 0);

   // server side - see man page for accept4 under linux 
   int socketfd = accept4( ... , SOCK_NONBLOCK);

It saves a little bit of work, but is less portable so I tend to set it with fcntl().