pm-utils does not follow requests to inhibit suspend/hibernate from scripts in /etc/pm/sleep.d

[Paulo J. S. Silva]

Binary package hint: pm-utils

1) ubuntu version

Description: Ubuntu 10.10
Release: 10.10

2) package version:

pm-utils:
  Instalado: 1.4.1-3
  Candidato: 1.4.1-3
  Tabela de versão:
 *** 1.4.1-3 0
        500 http://br.archive.ubuntu.com/ubuntu/ maverick/main amd64 Packages
        100 /var/lib/dpkg/status

3) If a script in /etc/pm/sleep.d returns a non-zero value upon requests to suspend or hibernate, the suspend/hibernate process should be aborted.

4) The system ignores the return value and supends/hibernates anyhow.

In order to reproduce the bug just drop the 00_avoid_suspend script in /etc/pm/sleep.d, make it executable and try to sleep.
The system should refuse to sleep but, at least here, it doesn't.

The problems seems to be in the _run_hook() function defined in file /usr/lib/pm-utils/pm-fuctions. This is the function that calls the scripts in the /etc/pm/sleep.d directory. After calling the script, _run_hook calls the function log twice before checking the script return value. It seems that the return value is being set back to 0 in these calls to log (but I am no bash expert).

To walk around I just added a line to store the script return value in a temporary variable and used that variable after the log calls to check the original result. Now it seems to work. My current _run_hook function looks like this:

_run_hook() {
 # $1 = hook to run
 # rest of args passed to hook unchanged.
 log "Running hook $*:"
 hook_ok "$1" && "$@"
 status=$?
 log ""
 log -n "$*: "
 hook_exit_status $status && LAST_HOOK="${1##*/}" || inhibit
}

Since I am no bash expert, I can not assure that this is a proper solution, but it works here.

[ariel cornejo (arielco)]

Created attachment 40891
patch for pm-functions.in

Since 1.4.0, _run_hook calls a log() after running the hook, and thus loses the hook exit status:

199 hook_ok "$1" && "$@"
200 log ""
201 log -n "$*: "
202 hook_exit_status $? && LAST_HOOK="${1##*/}" || inhibit

I attached a simple patch that seems to fix it.

This is important because e.g. a script that unmounts SD/MMC cards before suspending cannot inhibit it if the card is busy (see https://help.ubuntu.com/community/AspireOne/Ubuntu9.10?action=recall&rev=69).

Regards,
  Ariel

[ariel cornejo (arielco)]

Umm, I hoped there would be some kind of reply by now. Should I learn git and try to submit the patch there?

[Peter Wu (lekensteyn)]

hook_exit_status relies on the last exit code, but the function `log()` always returns 0.

The log lines adds something like:
"/usr/lib/pm-utils/sleep.d/000kernel-change hibernate hibernate:
and hook_exit_status:
 success.

Shouldn't the log lines be moved to hook_exit_status? Otherwise, the exitcode should be stored in a variable and applied to hook_exit_code as described in the bug report.

This bug does not occur at Debians pm-utils package, there are no log lines between hook_ok and hook_exit_code

[Peter Wu (lekensteyn)]

If we modify log() to make it return the previous error code, the following code are affected:

sleep.d/75modules
* resume_modules() calls modreload(), if it fails to load a module, the return value of sleep.d/75modules will match the return value of modprobe (is this the expected behavior, or should resume_modules return 0 just like suspend_modules? I vote for the latter one)

pm-functions
* load_hook_blacklist(): if disablehook() fails to write to ${STORAGEDIR}/disable_hook:HOOKNAME, the return value of load_hook_blacklist will be the return value of disablehook() (this is unlikely to happen, but should load_hook_)
* _run_hook: this function will work as expected as the return value of hook_ok propagates to hook_exit_status

Other functions/ scripts are not affected by this change in log() . Should log() change the previous exitcode or not?

The patch provided does not change the behavior of pm-utils, other than making _run_hook() work as it should do.

[Disidente (disidente)]

Had the same problem in xubuntu 11.04.

Published patch corrects pm-utils behaviour. I use a pm hook to umount SD cards on suspend, in order to avoid data corruption on resume (as described on https://bugs.launchpad.net/ubuntu/karmic/+source/linux/+bug/424877).

My hook looks like this:
#!/bin/sh
# Drop to: /etc/pm/sleep.d
# Use this script to prevent data loss on gnome-mounted MMC/SD
# cards. It syncs data and umounts all mmcblk devices prior to
# suspend, and cancels suspend if umounting was not posssible
# (i.e: something locks a file)
case "$1" in
 hibernate|suspend)
  sync
  for drive in $( ls /dev/mmcblk?p* ); do
      umount ${drive} > /dev/null
      # If umount failed: abort suspend
      if [ $? -gt 0 ]; then
   # Test if device keeps mounted. Previous command could fail
   # (i.e device was not mounted) with a non-stopper
   # problem for the suspend precess
   mount | grep ${drive}
   if [ $? -eq 0 ]; then
       echo "Failed to umount ${drive}" > /tmp/suspend.log
       exit 1
   fi
      fi
  done
  ;;
esac

My hook doesn't work without published patch, and without this hook data corruption is possible... so, I vote for high importance on this bug.

[Paulo J. S. Silva (pjssilva)]

Please, some one pass on the patch upstream. Or should I try to do it?

This bug have almost 8 months, and there is a patch that works. Please fix it. I got bitten by it once again after upgrading to 11.04...

[Peter Wu (lekensteyn)]

I've currently have the next command in my post-installation script for fixing the bug:

tmp="$(mktemp)"&&wget https://launchpadlibrarian.net/73767502/pm-functions-run-hooks-with-local.patch -O"$tmp"&&\
echo "8b59fdcb8f79c40510c15f6a8f3695f9569b084c6f6640e8175cf3105d81439e $tmp"|sha256sum -c&&\
sudo patch /usr/lib/pm-utils/pm-functions "$tmp";rm $tmp

The upstream bug has been reported two months after this one: https://bugs.freedesktop.org/show_bug.cgi?id=32210

[Brian Murray (brian-murray)]

With the patched version of pm-utils and the test case provided I see the following in /var/log/pm-suspend.log:

Running hook /etc/pm/sleep.d/05_avoid-sleep suspend suspend:

/etc/pm/sleep.d/05_avoid-sleep suspend suspend: Returned exit code 1.
Fri Jul 8 15:29:57 PDT 2011: Inhibit found, will not perform suspend

[Brian Murray (brian-murray)]

Attached is a debdiff that resolves this issue in Oneirc. It might be worth SRU'ing to stable releases of Ubuntu too.

[Launchpad Janitor (janitor)]

This bug was fixed in the package pm-utils - 1.4.1-8ubuntu1

---------------
pm-utils (1.4.1-8ubuntu1) oneiric; urgency=low

  * Add 26-inhibit-on-right-status.patch: Do not use the exit status of log
    rather the exit status of the hook thereby allowing inhibit to work.
    Thanks to Launchpad user Lekensteyn for the patch. (LP: #665651)
 -- Brian Murray <email address hidden> Fri, 08 Jul 2011 15:02:58 -0700

[Brian Murray (brian-murray)]

Well this ended up revealing an error in /etc/pm/sleep.d/novatel_3g_suspend provided by toshset and preventing suspend in Oneiric. So one should be extra cautious if SRU'ing this fix as it may break suspend and hibernate.