socket send

I got a little problem with a server / client, using TCP.

If I start my server, then start my client, all is good.
My client and server finds out if they "Loose connection", by checking returnvalue of recv().

But some of the functionality in my client use send() only, before any recv().
Meaning if I start my server, start my client, close the server, then my client will do send() to a closed connection.... And then I get segmentationFault.

How to avoid this?

What OS, and coudl you [post the relevant code?

OS is Redhat Linux.

The code failing:

Code:

---

void fail(void) {
  if(!(client->isConnected)) {
    fprintf(stderr,"Error, you are not connected to server\n");
    return;
  }

  safeSend(client->serverSock, "test", strlen("test")+1, client);
  safeSend(client->serverSock, "test", strlen("test")+1, client);
}
 
int safeSend(int incomingSocket, char *data, int  size, struct ClientInfo *client) {
  if(client->isConnected == 0) {
    fprintf(stderr,"RETURN -1\n");
    return -1;
  }

  int pos = 0;
  ssize_t status = 0;
  while(pos != size) {
    fprintf(stderr, "SEND IN FUNC\n");
    status = send(incomingSocket, &(data[pos]), size-pos, 0);
    fprintf(stderr, "END SEND IN FUNC\n");
    pos+=status;
    if(status == -1) {
     
      perror("send");
      return status;
    }
  }
  fprintf(stderr,"Return 0!\n");
  return 0; // All is well
}

---

What happens if I try to run fail() with an allready closed connection, is this:

Code:

---

SAFESEND:
SEND IN FUNC
END SEND IN FUNC
Return 0!
SEND IN FUNC

---

And there it just stops. And by stop, I mean that the program terminates.

you need to handle the return value before moving on to the rest of the code. I'm not sure on linux, but on windows you check the return value of send() and if it is SOCKET_ERROR (-1) you call WSAGetLastError() to get the actual cause of the error. In any case, once an error has occurred the socket is usually no longer valid.

So you set isConnected to false only when recv(2) returns 0? You should also set this when send(2) fails with EPIPE. Read man page.

Run this on your socket:

Code:

---

int set = 1;
setsockopt(incomingSocket, SOL_SOCKET, SO_NOSIGPIPE, (void *)&set, sizeof(set));

---

This will change the socket so it doesn't send the SIGPIPE signal when you attempt to write to a broken socket.

Another option is to install a signal handler for SIGPIPE. This way you wouldn't need to call setsockopt() on every socket.

On the opengroup-page for the recv() function, they specify that if it reads 0, it means the connection is lost.

But about send(), all it says is this:

Successful completion of a call to send() does not guarantee delivery of the message. A return value of -1 indicates only locally-detected errors.
Upon successful completion, send() shall return the number of bytes sent. Otherwise, -1 shall be returned and errno set to indicate the error.

Does this mean only recv() can neatly detect if the client has closed the connection?
I've seen people who check if send() return 0, in the same way as recv(), but it doesent work for me, and I can't find anything about that on the opengroup-pages :/

can u put both client and server code?

POSIX also writes:

Quote:

---

Originally Posted by Open Group

[EPIPE]
The socket is shut down for writing, or the socket is connection-mode and is no longer connected. In the latter case, and if the socket is of type SOCK_STREAM, the SIGPIPE signal is generated to the calling thread.

---

At least BSD supports MSG_NOSIGNAL in flags which tells send not to raise SIGPIPE.

If send(2) returns 0, it sent zero bytes if it conforms to standards. Check your system's man documentation.