Creating a TCP Ping Echo on User-Defined Protocols

Environment:Internet, networking

TCP Ping is necessary when measuring server functionality from different IP networks (mirror sites). As you see, it uses a TCP transport protocl instead of ICMP. To ping the remote (or local) server, it should be able to provide “NOOP” command (NO OPeration) followed by CRLF.

Such user-defined protocols can be:

• SMTP/25;
• FTP/21;
• POP3/110;
• and so forth (NOOP implemented).

Options included in demo project are the following:

• -cv: Connective ping (default)
• -ct: Continuous ping
• -n: Specify NOOP command
• -crlf: Append CRLF to NOOP
• -t<t>: Specify timeout

Hence, you can specify NOOP command yourself, of course, if it differs from the default.

The reasonable question: How do you ping the HTTP server? To ping the HTTP server, you should specify a NOOP-command (because it’s not implelemented in HTTP/1.x) as the implemented method. This time, a connective ping is better than a continuous ping. Let’s see the connective ping source:

unsigned int CPing::PingConnective(char* szHost,
                                   unsigned int iPort,
                                   unsigned int iPackets)
{
  struct hostent* host = NULL;
  struct sockaddr_in saddr;
  unsigned int s = 0;
  unsigned int dw1, dw2, dw3;
  char szBuffer[256];

  if (iPackets>MAX_SENDS) return (0);
  free (Res);
  Res = (pingstore*)malloc (sizeof(pingstore)*iPackets);
  memset (Res, 0, sizeof(pingstore)*iPackets);
  if (!iBytesToRecv) iBytesToRecv = strlen(szNoop);

  host = gethostbyname (szHost);
  if (host==NULL) return (0);
  saddr.sin_family = AF_INET;
  saddr.sin_port = htons(iPort);
  saddr.sin_addr = *((struct in_addr*)host->h_addr);

  for (int i=0;i< iPackets;i++)
  {
    s = socket(AF_INET, SOCK_STREAM, 0);
    if (!s) return ((iTotalRes)?1:0);
    setsockopt(s, SOL_SOCKET, SO_RCVTIMEO,
              (char*)&iTimeOut, sizeof(iTimeOut));
    setsockopt(s, SOL_SOCKET, SO_SNDTIMEO,
              (char*)&iTimeOut, sizeof(iTimeOut));
    if (connect (s,(struct sockaddr*)&saddr, sizeof(saddr))
        == -1) return ((iTotalRes)?1:0);

    iTotalRes++;
    sprintf (szBuffer, "%srn", szNoop);

    dw1 = GetTickCount();
    int iSent = send (s, szBuffer, strlen(szBuffer), 0);
    dw2 = GetTickCount();
    int iRecv = recv (s, szBuffer, iBytesToRecv, 0);
    dw3 = GetTickCount();

    Res[i].iPort       = iPort;
    Res[i].iTimeSend   = dw2-dw1;
    Res[i].iTimeRecv   = dw3-dw2;
    Res[i].iTotalSent  = ((iSent==SOCKET_ERROR)?0:iSent);
    Res[i].iTotalRecvd = ((iRecv==SOCKET_ERROR)?0:iRecv);

    closesocket (s);
  }
  return (1);
}

You can use a simple Perl script for it:

#!/usr/local/bin/perl
# ################################
# Connective ping to http-server
# ################################

die("httpping.pl <host> <packets> [-p<port>]n")
     unless (scalar(@ARGV)>=2);
$host    = $ARGV[0];
$port    = 80;    # default, but you can change it
$packets = $ARGV[1];
$noop    = "GET / HTTP/1.0";

for ($i=2; $i< scalar(@ARGV);$i++)
{
  if ($ARGV[$i] =~ /^(x2Dx70)/)
  {
    $port = substr ($ARGV[$i],2,length($ARGV[$i])-2);
  }
}

open (PING, "|ping.exe $host $port $packets -n -cv -crlf") ||
              die ("Error: ping executable not foundn");
print PING "$noopn";
close (PING);

Example of Usage

> ping localhost 8080 10 -n -crlf -cv    # connective ping
Enter valid NOOP command: GET / HTTP/1.0
...

In conclusion, note that you can also modify the timeout value to extend or lower the “live time” of requests. I hope this will be very helpful for system administrators to audit their systems on “request-response” ability.

Downloads

Download demo program – 24 Kb
Download source – 2 Kb