Temporarily increasing sudo's timeout for the duration of an install script

I'm trying to write a script that will install a bunch of software and I'd like to not have to run everything as root, so I'd like to be able to prompt for a password and then go about the install, using sudo or su to get privileges when I need them.

I was doing a sudo -v to prompt for a password at the beginning of the script, and then just using sudo normally later on. This works great until I get to a single install that takes over the timeout.

I'd rather not have to permanently increase the timeout. Is there a way I can increase sudo's timeout for the current session only?

Solution 1:

You can setup a loop that runs in the background to periodically execute "sudo -v", the trick of course is getting the loop to cleanly terminate when your script terminates. So there has to be some type of communication between the two processes; tmp files are fine for this, and they can easily be cleaned up after the script runs, too. (An install script usually does this, anyway.)

For example (remove the 'echo' statements to use this; these just show it "working"):

#!/bin/bash
log=running_setup.txt
sudo_stat=sudo_status.txt

echo "========= running script $$ ========"
echo $$ >> $sudo_stat
trap 'rm -f $sudo_stat >/dev/null 2>&1' 0
trap "exit 2" 1 2 3 15

sudo_me() {
 while [ -f $sudo_stat ]; do
  echo "checking $$ ...$(date)"
  sudo -v
  sleep 5
 done &
}


echo "=setting up sudo heartbeat="
sudo -v
sudo_me

echo "=running setup=" | tee $log
while [ -f $log ]
do
 echo "running setup $$ ...$(date) ===" | tee -a $log
 sleep 2
done

# finish sudo loop
rm $sudo_stat

Then you'll see... (note: the pid is put into the tmp file, just so you can easily kill it. It's not necessary, though):

$ ./do_it.sh
========= running script 6776 ========
=setting up sudo heartbeat=
[sudo] password for user: 
=running setup=
checking 6776 ...Wed May  4 16:31:47 PDT 2011
running setup 6776 ...Wed May  4 16:31:48 PDT 2011 ===
running setup 6776 ...Wed May  4 16:31:50 PDT 2011 ===
running setup 6776 ...Wed May  4 16:31:52 PDT 2011 ===
checking 6776 ...Wed May  4 16:31:53 PDT 2011
running setup 6776 ...Wed May  4 16:31:54 PDT 2011 ===
<ctrl-c>  (cleans up files, then exits)

Solution 2:

I liked michael_n's answer, but had the most irrational desire not to use a temp file. Maybe this can provide some perspective.

My solution was:

#!/bin/bash
function sudo_ping() {
    if [[ ! -z $SUDO_PID ]]; then
        if [[ $1 -eq stop ]]; then
            echo "Stopping sudo ping in PID = $SUDO_PID"
            kill $SUDO_PID
            return
        else
            echo "Already sudo pinging in PID = $SUDO_PID"
            return
        fi
    fi

    echo "Starting background sudo ping..."
    sudo -v
    if [[ $? -eq 1 ]]; then
        echo "Oops, wrong password."
        return
    fi
    sudo echo "ok"

    while true; do
        echo 'Sudo ping!'
        sudo -v
        sleep 1
    done &
    SUDO_PID=$!
    sudo echo "Sudo pinging in PID = $SUDO_PID"

    # Make sure we don't orphan our pinger
    trap "sudo_ping stop" 0
    trap "exit 2" 1 2 3 15
}

sudo_ping
sleep 5
echo "Goodbye!"

Again, the echo's are extraneous...

$ ./sudoping.sh 
Starting background sudo ping...
Password:
ok  
Sudo ping!
Sudo pinging in PID = 47531
Sudo ping!
Sudo ping!
Sudo ping!
Sudo ping!
Goodbye!
Stopping sudo ping in PID = 47531

Again, ctrl-c works too...

$ ./sudoping.sh 
Starting background sudo ping...
ok  
Sudo ping!
Sudo pinging in PID = 47599
Sudo ping!
^CStopping sudo ping in PID = 47599

Solution 3:

Based on this gist, I've made a concise and clean version:

# Prevent sudo timeout
sudo -v # ask for sudo password up-front
while true; do
  # Update user's timestamp without running a command
  sudo -nv; sleep 1m
  # Exit when the parent process is not running any more. In fact this loop
  # would be killed anyway after being an orphan(when the parent process
  # exits). But this ensures that and probably exit sooner.
  kill -0 $$ 2>/dev/null || exit
done &