Wait command doesn't work as expected with init script and pidfile

[sebastian_de]

I'm trying to create an init script that can shut down a VM and waits for the shutdown to complete, before shutting down OpenWrt. I found out that the wait command doesn't work as expected when using it in an init script.

This script works as expected:

#!/bin/sh

sleep_pidfile="/var/run/sleep2.pid"

sleep 5s &
echo $! > $sleep_pidfile

echo "PID is: $(cat $sleep_pidfile)"
wait $(cat $sleep_pidfile)

echo "wait exited with code: $?"
rm $sleep_pidfile

It waits 5 seconds and the exit code is 0:

# ./pidfiletest.sh 
PID is: 5942
wait exited with code: 0

With an init script the same thing doesn't work, the script exits immediately and wait exits with 127:

#!/bin/sh /etc/rc.common

START=99
STOP=1

sleep_pidfile="/var/run/sleep.pid"

start() {
sleep 300 &

echo $! > $sleep_pidfile
echo "sleep PID is: $(cat $sleep_pidfile)"
} 

stop() {
echo "Executing insomnia.sh in background"
/root/insomnia.sh &

echo "waiting for process with PID: $(cat $sleep_pidfile)"
wait $(cat $sleep_pidfile)
echo "wait exit code: $?"
rm $sleep_pidfile
}

# /etc/init.d/rc_pidtest start
sleep PID is: 6018

# /etc/init.d/rc_pidtest stop
waiting for process with PID: 6018
wait exit code: 127

Also running wait manually with the PID doesn't work:

# wait 6018
# echo $?
127

Does anybody know what's causing this?

[jeff]

Just ran a quick check, as wait is a shell built-in and who knows just what busybox implements. It seems to be working on a recent openwrt-18.06 build (I don't recall any changes to busybox config related to wait)

jeff@office:~$ sh -c 'sleep 20' & 
jeff@office:~$ ps
  PID TTY          TIME CMD
 5566 pts/0    00:00:00 ash
 5583 pts/0    00:00:00 sleep
 5584 pts/0    00:00:00 ps
jeff@office:~$ wait 5583 && echo $?
0
[1]+  Done                       sh -c "sleep 20"

Maybe try running your script with -x to get some insight?

[sebastian_de]

I forgot to mention, I'm on branch openwrt-18.06 as well, busybox is v1.28.3.
On my machine, your test works as expected, too. My first thought was, that the pidfile handling in my init script was wrong, so I wrote the simpler script (the first example). The weird thing is that they do more or less the same. The simple script works, the init script doesn't.

# ps
 7127 root      3084 S    sleep 300

# sh -x /etc/rc.common /etc/init.d/rc_pidtest stop
+ . /lib/functions.sh
+ N='
'
+ _C=0
+ NO_EXPORT=1
+ LOAD_STATE=1
+ LIST_SEP=' '
+ reset_cb
+ '[' -z  -a -f /lib/config/uci.sh ]
+ . /lib/config/uci.sh
+ CONFIG_APPEND=
+ . /lib/functions/service.sh
+ SERVICE_QUIET=1
+ SERVICE_SIG_RELOAD=HUP
+ SERVICE_SIG_STOP=TERM
+ SERVICE_STOP_TIME=5
+ SERVICE_MATCH_EXEC=1
+ initscript=/etc/init.d/rc_pidtest
+ action=stop
+ shift 2
+ 
+ . /etc/init.d/rc_pidtest
+ START=99
+ STOP=1
+ sleep_pidfile=/var/run/sleep.pid
+ '[' -n  ]
+ ALL_COMMANDS='start stop reload restart boot shutdown enable disable enabled depends '
+ list_contains ALL_COMMANDS stop
+ local 'var=ALL_COMMANDS'
+ local 'str=stop'
+ local val
+ eval 'val=" ${ALL_COMMANDS} "'
+ val=' start stop reload restart boot shutdown enable disable enabled depends  '
+ '[' ' start' '!=' ' start stop reload restart boot shutdown enable disable enabled depends  ' ]
+ stop
+ cat /var/run/sleep.pid
+ echo 'waiting for process with PID: 7127'
waiting for process with PID: 7127
+ cat /var/run/sleep.pid
+ wait 7127
+ echo 'wait exit code: 127'
wait exit code: 127
+ rm /var/run/sleep.pid

So the script correctly executes wait 7127. As I wrote before: running wait 7127 manually doesn't work either.

Maybe I'll try to install bash and see if the same happens there.

[sebastian_de]

Behavior on bash stays the same, but the error message is more helpful:

# wait 8202
bash: wait: pid 8202 is not a child of this shell

So I found the cause of this error - I guess I have to find another solution instead of wait
But thanks for your response!

[sebastian_de]

WAIT for "any process" to finish

This stackoverflow has some solutions for this problem. I went with

while [ -e /proc/$(cat $sleep_pidfile) ]; do sleep 1s; done

which works