my puppy won't always go to sleep

[Krzys Majewski <majewski@cs.ubc.ca>]

I've found a  way of putting my machine to sleep  (so nice, don't have
to  boot  anymore  just  to  check  my  mail..),  but  it's  not  very
reliable. Sometimes I have to run the script two or three times before
everything shuts down. It looks like the hard drive occasionally spins
up again  at the same time that  apm --suspend is called.  I thought I
could  prevent this by  calling apm  (without the  --suspend argument)
earlier in the script in order to "cache" the executable in RAM, but I
guess this doesn't  really do the trick. Here's  the script, maybe y'all
can give me your  $0.02. BTW I do know about apmd,  I'm just not using
it because  when I tried,  weird things happened (don't  remember what
they  are,  sorry). Though  if  you  are  using apmd  successfully  to
accomplish what I'm  trying to here, feel free to let  me know how you
did it. -chris

#!/bin/sh

PRGNAME=`basename $0`
RUNLEVELFILE=/var/local/${PRGNAME}/runlevel
runlevel | awk '{print $2}' > ${RUNLEVELFILE}
ATTEMPTS=5  
COUNT=0
HDD=/dev/hda

stop-processes # send STOP signal to things that need networking
sudo /sbin/telinit 4 # this shuts down networking, noflushd, and gpm
while [ ${COUNT} -lt ${ATTEMPTS} ]; do # keep trying hdparm -y until success
    sync
    sleep 1
    sync
    sleep 1
    sync
    sleep 1
    apm >> /dev/null
    hdparm ${HDD} >> /dev/null
    grep >> /dev/null 2>&1
    hdparm -y ${HDD} >> /dev/null
    sleep 5
    if ! hdparm -C ${HDD} | grep active >> /dev/null; then
	break
    fi
    COUNT=`expr ${COUNT} + 1`
done
apm --suspend 
sudo -u root /sbin/telinit `cat ${RUNLEVELFILE}` # resume original runlevel
resume-processes # send CONT signal to processes we had sent STOP to