In article ,
"Nuertey Odzeyem" wrote:

> Hello,
> I have a device that is running Interpeak IPNET - Dual IPv4/IPv6 Stack on
> OSE. Part of my stack test involves running essentially the perl script
> below numerous times. Notice that the last but one line of my script is
> deliberately commented out. I wanted to test an ungraceful exit case.
>
> $telnet = new Net::Telnet ( Timeout=>10, Errmode=>'die');
> $telnet->open('10.8.118.215');
> $telnet->login($login, $password);
>
> # No 'exit' for the Telnet session since I am testing the UNGRACEFUL
> exit case
> # $telnet->print("exit\r\n");
> $telnet->close();
>
> I discovered that my device leaks memory for each of these telnet sessions
> and that it eventually resets after about 200 sessions. Debugging the
> source, I see that because I do not exit the session (see commented out
> line) it takes about 30 seconds before the server detects a "connection
> reset by peer". Then when the server kills the proxy process, the memory for
> it is somehow not released. So I have two problems: (1) how can I get the
> server to detect the lost connection immediately; (2) Why is the memory for
> the proxy processes not being released even though the processes themselves
> are eventually killed.

This is a bug in the device. Report it to the device's manufacturer,
and maybe they'll fix it in a firmware update.

>
> Note that when I uncomment out the "$telnet->print("exit\r\n");" the closed
> connection is detected immediately (obviously it would since we did close
> it), the process is killed and the memory released properly. The puzzling
> thing is it is the same code segment that is executed in both cases. Only
> the entry points differ--INET_ERRNO as opposed to CHILD_DONE
>
> Could I do something with SO_KEEPALIVE or SO_LINGER here? Is my testcase
> even valid? What does the "close" do without the "exit"?

Why don't you just do the "exit"?

--
Barry Margolin, barmar@alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***
*** PLEASE don't copy me on replies, I'll read them in the group ***

On Jul 3, 1:09 pm, "Nuertey Odzeyem"
wrote:

> Could I do something with SO_KEEPALIVE or SO_LINGER here?

No.

> Is my testcase
> even valid?

Yes. It proves the device is broken/buggy.

> What does the "close" do without the "exit"?

It shuts down the TCP connection without terminating the application
on the other side first. This is a common case and I think it's fair
to argue that a device that mishandles it is broken.

DS