Delayed kernel panic in Linux

For our tests we need to trigger kernel panic on a remote server (VM), via SSH:

ssh server "echo c > /proc/sysrq-trigger"

The problem is that in most cases the SSH session gets stuck, since the kernel panic happens before the connection is disconnected. There is a generic connection timeout, but this is not good enough.

Is there a way to delay the panic?

We tried to put the following in a file on the server:

# - panic.sh -

#/bin/bash
sleep 5
echo c > /proc/sysrq-trigger

And then to execute it:

ssh server "nohup panic.sh &"

But this didn't help. The SSH session waits till the sleep ends.

Solution

This happens because your script keeps all pipes open, so ssh has to wait to see if you'll write more.

If you close them all, ssh knows it won't receive any more output and will exit:

$ time ssh localhost 'sleep 5 < /dev/null > log 2>&1 &'
real    0m0.171s
user    0m0.013s
sys     0m0.003s

You can also do this from within the script using the exec command before the sleep:

exec < /dev/null > log 2>&1