[SOLVED] bash script (qmail) won't boot but starts from console (debian lenny)

dbj · 11-17-2009, 07:49 PM

In /etc/rc2.d/ I have a link to
/etc/init.d/qmail

It will not start when I boot (xmalloc error : ../bash/subst.c cannot allocate ...),
but I can on the command line do

/etc/init.d/qmail start

and

sh /etc/init.d/qmail start

but not
. /etc/init.d/qmail start

The latter gives the same error as during boot

How can I work around it?
Other scripts work.

dbj · 11-17-2009, 09:14 PM

I killed the S20qmail in /etc/rc2.d and instead,
I added the following to /etc/rc.local

export SHELL=/bin/bash
sh /etc/init.d/qmail start
exit 0

That helped. I have also tried to write "export SHELL=/bin/bash" at the beginning of /etc/init.d/qmail but somehow I couldn't get it to work

catkin · 11-18-2009, 01:01 AM

Intriguing. Can you post your /etc/init.d/qmail?

dbj · 11-18-2009, 01:40 AM

It successfully starts qmail-start but afterwards something goes wrong

Code:

#!/bin/bash
#
# /etc/init.d/qmail : start or stop the qmail mail subsystem.
#
# Written by Christian Hudon <chrish@debian.org>
# Currently maintained by Jon Marler <jmarler@debian.org>
#
# Configuration
#


# set default delivery method

alias_empty="|/usr/sbin/qmail-procmail"  # procmail delivery to /var/spool/mail
#alias_empty="./Maildir/"       # This uses qmail prefered ~/Maildir/ directory
				# You may want to maildirmake /etc/skel/Maildir
#alias_empty="./Mailbox"        # This uses Mailbox file in users $HOME

logger="splogger qmail 2"	# facility mail == 2
#logger="|accustamp >>/var/log/qmail.log"   # If you have accustamp installed.
#logger=">>/var/log/qmail.log"              # Does not give timing info.

# If you uncommented one of the lines that appends to /var/log/qmail.log, you
# need to uncomment the following two lines.
#touch /var/log/qmail.log
#chown qmaill /var/log/qmail.log

# If you want to use one or more of the Relay Black Lists, uncomment
# the appropriate lines.

rblmsg=
rblsmtpd=
#rblmsg=" (with rblsmtpd)" 
#rblsmtpd="/usr/bin/rblsmtpd -r list.dsbl.org -r relays.ordb.org"

#
# End of configuration
#

test -x /usr/sbin/qmail-start || exit 0
test -x /usr/sbin/qmail-send || exit 0

case "$1" in
    start)
	echo -n "Starting mail-transfer agent: qmail" $rblmsg
	sh -c "start-stop-daemon --start --quiet --user qmails \
		 --exec /usr/sbin/qmail-send \
		 --startas /usr/sbin/qmail-start -- \"$alias_empty\" $logger &"
	# prevent denial-of-service attacks, with ulimit
	ulimit -v 16384
         sh -c "start-stop-daemon --start --quiet --user qmaild \
	    --pidfile /var/run/tcpserver_smtpd.pid --make-pidfile \
            --exec /usr/bin/tcpserver -- -R -H \
            -u `id -u qmaild` -g `id -g nobody` -x /etc/tcp.smtp.cdb 0 smtp \
            $rblsmtpd /usr/sbin/qmail-smtpd 2>&1 \
            | $logger &"

	# Uncomment the following lines to automatically start the pop3 server
	#sh -c "start-stop-daemon --start --quiet --user root \
	#    --pidfile /var/run/tcpserver_pop3d.pid --make-pidfile \
	#    --exec /usr/bin/tcpserver -- -R -H \
	#    0 pop-3 /usr/sbin/qmail-popup `hostname`.`dnsdomainname` \
	#    /usr/bin/checkpassword /usr/sbin/qmail-pop3d Maildir &"
	
	echo "."
	;;
    stop)
	echo -n "Stopping mail-transfer agent: qmail" $rblmsg
	if [ "`pidof /usr/sbin/qmail-send`" ] ; then
	    start-stop-daemon --user qmails --stop --quiet --oknodo --exec /usr/sbin/qmail-send
	    start-stop-daemon --user qmaild --stop --quiet --oknodo --pidfile /var/run/tcpserver_smtpd.pid --exec /usr/bin/tcpserver 
	    # Uncomment the following line if you have enabled the pop3 server
	    #start-stop-daemon --user root --stop --quiet --oknodo --pidfile /var/run/tcpserver_pop3d.pid --exec /usr/bin/tcpserver

	    # Wait until the timeout for qmail processes to die.
	    count=120
	    numdots=0
	    while ([ $count != 0 ]) do
		let count=$count-1
		if [ "`pidof /usr/sbin/qmail-send`" ] ; then
		    echo -n .
		    let numdots=$numdots+1
		    sleep 1
		else
		    count=0
		fi
	    done

	    # If it's not dead yet, kill it.
#	    if [ "`pidof /usr/sbin/qmail-send`" ] ; then
#		echo -n " TIMEOUT!"
#		kill -KILL `pidof /usr/sbin/qmail-send`
#	    else
		case $numdots in
		  0) echo "." ;;
		  1) echo ;;
		  *) echo " done." ;;
		esac
#	    fi
	else
	    echo " not running.";
	fi

	;;
    restart)
	$0 stop
	$0 start
	;;
    cdb)
	echo "Rebuilding tcp.smtp.cdb."
	cd /etc
	tcprules tcp.smtp.cdb tcp.smtp.temp < tcp.smtp
	;;
    flush)
        /usr/sbin/qmail-tcpok
        start-stop-daemon --stop --quiet --oknodo --signal ALRM --exec /usr/sbin/qmail-send
        echo "Queue flushed."
        ;;
    stat)
        /usr/sbin/qmail-qread
        /usr/sbin/qmail-qstat
        ;;
    reload|force-reload)
	echo "Reloading 'locals' and 'virtualdomains' control files."
	start-stop-daemon --stop --quiet --oknodo --signal HUP --exec /usr/sbin/qmail-send
	;;
    *)
	echo 'Usage: /etc/init.d/qmail {start|stop|stat|cdb|restart|reload}'
	exit 1
esac

exit 0

catkin · 11-18-2009, 08:19 AM

Try adding an echo above and below the ulimit command so you can see which of the three significant commands in the start section it is barfing on; probably the ulimit but let's make sure before going further.

dbj · 11-18-2009, 09:03 AM

Quote:

Originally Posted by catkin

Try adding an echo above and below the ulimit command so you can see which of the three significant commands in the start section it is barfing on; probably the ulimit but let's make sure before going further.

Tried that and it is not ulimit. It happens when the start-stop-daemon tries to start the tcpserver.
I have reason to believe that start-stop-daemon might be buggy or not compatible with lenny.

catkin · 11-18-2009, 10:37 AM

Quote:

Originally Posted by dbj

I have reason to believe that start-stop-daemon might be buggy or not compatible with lenny.

Care to share the reason with us?

dbj · 11-18-2009, 11:31 AM

Quote:

Originally Posted by catkin

Care to share the reason with us?

Alright, it is a little bit complicated. It is not fair to say that start-stop-daemon is buggy, but we have a conflict.
I have also tried to set it up using spamdyke like

Code:

...
         rblsmtpd="/usr/local/bin/spamdyke -l -f /etc/spamdyke.conf"
...
...
         sh -c "start-stop-daemon --start --quiet --user qmaild \
            --pidfile /var/run/tcpserver_smtpd.pid --make-pidfile \
            --exec /usr/bin/tcpserver -- -R -H \
            -u `id -u qmaild` -g 1004 -x /etc/tcp.smtp.cdb 0 smtp \
             $rblsmtpd /usr/sbin/qmail-smtpd 2>&1 \
            | $logger &"

It works I can send emails (if I use my rc.local hack), but every time a mail is sent I get in /var/log/mail.err

server1 spamdyke[2200]: ERROR: unable to find protocol number with getprotobyname()

That is ridiculous. The code in spamdyke is very simple it just call
getprotobyname("udp"), and if the returned pointer is null it logs the error message. My guess is that getprotobyname for some reason hasn't been able to read the /etc/protocols.
I have made a small testprogram using getprotobyname and it works fine. I even let start-stop-daemon start my test program.
The testprogram was build with the same settings as spamdyke. Afterwards I made a clean and make spamdyke, and a reinstallation but it changed nothing.
I have

server1:~# dpkg -l | grep ucspi
ii ucspi-tcp 1:0.88-2 command-line tools for building TCP client-server applications

ucspi was installed twice first 0.88-15 then the above.

I tried to run tcpserver from the command line ie. not as a daemon and suddenly I got no getprotobyname error

In other words tcpserver likes my console enviroment, but not when it is invoked by something else.
I must add that I have used the above configuration on an etch without problems.
It is possible that it is caused by two different errors, but I think it is start-stop-daemon/tcpserver/lenny conflict. Maybe I should add that I use AMD64 and have a raid filesystem.

catkin · 11-18-2009, 11:40 AM

Quote:

Originally Posted by dbj

Alright, it is a little bit complicated.

Can't argue with that!

Not my area of expertise (haven't even heard of spamdyke) so I can't help much (worth another thread with a more specific subject?) but what is the connection between the error message in the OP (xmalloc error : ../bash/subst.c cannot allocate ...) which looks like a bash problem and what you described in the last post which seems unrelated?! An incompatible library somewhere triggering both seemingly unrelated errors?

dbj · 11-21-2009, 07:03 AM

ulimit is simply too low