wall hangs on tty_write system call for about 15 minutes on debian 2.6 kernel

[KRUPASINDHU PRADHAN]

--------------------------------------------------------------------------------

Hi
My system was about 20% cpu utilization. Console was open and some other terminals were oepn.

0. system info Linux Node1 2.6.10-telco-1.35.1-mckinley-smp #1 SMP Sun Nov 18 20:58:46 UTC 2007 ia64 GNU/Linux
1. System had invoked shutdown -r now commmand
2. As part of it , it was putting wall message to all ttys
The wall message is "The system is going down for reboot NOW!"
3. We found that wall was hanging for 15mins and was not posting this message to tty.
4. Because of above reason, shutdown was hanging for about 15 minutes.


===Code===
system ("while ( true );do sleep 25;date >> $rnps;ps -emo pid,ppid,psr,pcpu,pmem,vsz,rss,lwp,state,time,wchan=WIDE-WIDE-WCHAN-COLUMN,args>> $rnps; echo >> $rnps;done &");
...
..
shutdown -r now

====Code Ends===

Below is seen for about 15-mins as captured from ps -e
30133 30131 - 0.0 0.0 3632 1744 - - 00:00:00 -/usr/bin/wall
- - 3 0.0 - - - 30133 S 00:00:00 tty_write


Thanks
Krupa

[KRUPASINDHU PRADHAN]

**ANY ONE HELPING ME ON THIS ISSUE***

[Matti_Kurkela · Accepted solution]

Terminals? Do you mean you're using terminals attached with serial cables?

A serial-attached TTY might have the serial port handshaking lines stuck in a configuration that stops the server from sending any data to the terminal.

Of course, the wall command should be able to skip this automatically. But I think the use of serial-attached terminals (other than the console) is so rare these days, I suspect the wall command has not been tested against common failure modes of serial-attached terminals.

You might want to file a bug report to the manufacturer of your Linux distribution.

MK

[KRUPASINDHU PRADHAN]

Hi.
wall does not hang all the time. Your concern over tty seriel port handshaking stuck might have caused this to hang. Should not it hang all the time, in that case ?

One clear observation is that it hangs for "15-minutes". And the system call at which it hung was tty_write.

Note we have ssh sessions open from another system to this system. The ssh sessions are closed, before this wall was invoked. Those ssh sessions are from logical IPs. Does this cause system to hang. I read a topic on ssh tunnel hung case.

this hung session timeout in 15-minutes, explanation given from ssh-tunnel http://www.jfranken.de/homepages/johannes/vortraege/ssh2_inhalt.en.html

Do you think, our tty_write could hang because of above ssh session closed.