unable to shut down the system after suspend / resume

Bug #1594035 reported by Donald Pellegrino on 2016-06-18
This bug affects 4 people
Affects Status Importance Assigned to Milestone
systemd (Ubuntu)

Bug Description

In the case of a system that has gone through a suspend / resume cycle, the "shutdown now" command fails with the following output:

$ shutdown now
Failed to power off system via logind: Transaction is destructive.
Failed to start poweroff.target: Interactive authentication required.
See system logs and 'systemctl status poweroff.target' for details.
Failed to open /dev/initctl: Permission denied
Failed to talk to init daemon.


$ systemctl status poweroff.target
โ— poweroff.target - Power-Off
   Loaded: loaded (/lib/systemd/system/poweroff.target; disabled; vendor preset:
   Active: inactive (dead)
     Docs: man:systemd.special(7)


$ systemctl list-jobs
438 dev-mapper-cryptswap1.device start running
439 dev-mapper-cryptswap1.swap start waiting
436 <email address hidden> start running

3 jobs listed.


Steps to reproduce:

1. Boot system
2. User login to graphical session
3. Close lid while on battery power triggering a suspend
4. Open lid while on battery power triggering a resume
5. Run "shutdown now" from a terminal. Note that the command must be run from the terminal to see output. Selecting "Shut Down..." from the system menu in the menu bar will hide the error output.
6. Observe that the system fails to shut down

Expected behavior:

At step 6, the system should shut down gracefully.

Note that the shutdown process works as expected if steps 3 and 4 are skipped. I suspect that something about the suspend / resume cycle during the session has an interaction effect with the shutdown process, preventing it from executing correctly.

ProblemType: Bug
DistroRelease: Ubuntu 16.04
Package: systemd 229-4ubuntu6
ProcVersionSignature: Ubuntu 4.4.0-24.43-generic 4.4.10
Uname: Linux 4.4.0-24-generic x86_64
ApportVersion: 2.20.1-0ubuntu2.1
Architecture: amd64
CurrentDesktop: Unity
Date: Sat Jun 18 18:04:26 2016
EcryptfsInUse: Yes
 Bus 002 Device 001: ID 1d6b:0003 Linux Foundation 3.0 root hub
 Bus 001 Device 003: ID 04f2:b550 Chicony Electronics Co., Ltd
 Bus 001 Device 002: ID 8087:0a2a Intel Corp.
 Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
MachineType: System76, Inc. Lemur
 PATH=(custom, no user)
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-4.4.0-24-generic.efi.signed root=UUID=7aecd352-8a3f-4ce9-88b8-0fc048dc9660 ro quiet splash vt.handoff=7
SourcePackage: systemd
UpgradeStatus: No upgrade log present (probably fresh install)
dmi.bios.date: 11/29/2015
dmi.bios.vendor: American Megatrends Inc.
dmi.bios.version: 1.05.06RS76
dmi.board.asset.tag: Tag 12345
dmi.board.name: Lemur
dmi.board.vendor: System76, Inc.
dmi.board.version: lemu6
dmi.chassis.asset.tag: No Asset Tag
dmi.chassis.type: 10
dmi.chassis.vendor: System76, Inc.
dmi.chassis.version: N/A
dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvr1.05.06RS76:bd11/29/2015:svnSystem76,Inc.:pnLemur:pvrlemu6:rvnSystem76,Inc.:rnLemur:rvrlemu6:cvnSystem76,Inc.:ct10:cvrN/A:
dmi.product.name: Lemur
dmi.product.version: lemu6
dmi.sys.vendor: System76, Inc.

I confirmed that disabling encrypted swap is a work-around for the inability to shut down. I commented out the cryptswap1 entry from /etc/crypttab:

$ diff /etc/crypttab /etc/crypttab~
< # cryptswap1 UUID=d1bd47b1-9969-4cf6-bb7e-482a50b41ffa /dev/urandom swap,offset=1024,cipher=aes-xts-plain64
> cryptswap1 UUID=d1bd47b1-9969-4cf6-bb7e-482a50b41ffa /dev/urandom swap,offset=1024,cipher=aes-xts-plain64

I also set /etc/fstab to use the swap partition directly rather than /dev/mapper/cryptswap1:

$ diff /etc/fstab /etc/fstab~
< UUID=d1bd47b1-9969-4cf6-bb7e-482a50b41ffa none swap sw 0 0
< #/dev/mapper/cryptswap1 none swap sw 0 0
> #UUID=d1bd47b1-9969-4cf6-bb7e-482a50b41ffa none swap sw 0 0
> /dev/mapper/cryptswap1 none swap sw 0 0

On this machine, it seems encrypted swap was not working anyway as per lines in the attached JournalErrors.txt:

Jun 18 17:56:14 username-Lemur systemd-cryptsetup[446]: Failed to activate with key file '/dev/urandom': Device or resource busy
Jun 18 17:56:18 username-Lemur systemd-cryptsetup[446]: Loading of cryptographic parameters failed: Invalid argument
Jun 18 17:56:18 username-Lemur systemd-cryptsetup[446]: Failed to activate: Invalid argument
Jun 18 17:56:18 username-Lemur systemd[1]: Failed to start Cryptography Setup for cryptswap1.
Jun 18 17:56:18 username-Lemur systemd[1]: Dependency failed for Encrypted Volumes.
Jun 18 17:56:18 username-Lemur systemd[1]: Dependency failed for dev-mapper-cryptswap1.device.
Jun 18 17:56:18 username-Lemur systemd[1]: Dependency failed for /dev/mapper/cryptswap1.
Jun 18 17:56:18 username-Lemur systemd[1]: Dependency failed for Swap.
Jun 18 17:56:18 username-Lemur systemd[1]: <email address hidden>: Failed with result 'exit-code'.
Jun 18 17:56:18 username-Lemur systemd-cryptsetup[805]: Failed to activate with key file '/dev/urandom': Device or resource busy

With encrypted swap disabled as per the changes to /etc/crypttab and /etc/fstab, the machine now shuts down as expected even when suspend/resume cycles occur during the user session.

Bug #1447282 may be relevant.

Loye Young (loyeyoung) wrote :

Being a fellow pilgrim in the Way of the Penguin, I can confirm the exact same facts as Mr. Pellegrino on clean install of Ubuntu Mate 16.04.

It appears that the swap partition is not actually encrypted at all. Syslog shows that encryption failed, and "cryptsetup -v isLuks /path/to/partition" shows not LUKS partition. This is so no matter which path to the swap partition I use, including: /dev/disks/by-uuid/XXXX, the device show by "grep /proc/swaps", or /dev/mapper/cryptswap1.

Looking at /var/log/syslog, I see that cryptsetup failed because /dev/urandom is not available. ("grep crypt /var/log/syslog" for details.)

Further, I notice that poweroff.target is disabled. When I enable it (systemctl enable poweroff.target), shutdown works as expected unless the computer has resumed from suspend.

The work around suggested by Mr. Pellegrino works, but of course that means that swap is not encrypted, which is of course a security vulnerability.

Here is my working theory: On boot-up, systemd tries to create an encrypted swap, but when it cannot, systemd creates an unencrypted swap. (Feature or bug? There would be competing considerations, so it is hard to say.) After resume from suspend, which of course involves (on suspend) writing RAM to swap and then (on resume) reading from swap to RAM, the system thinks there should be an encrypted swap (because that's what /etc/fstab and /etc/crypttab say), but can't find it and gets confused when time comes to shutdown.

This being a security issue, it should be given attention.

Changed in systemd (Ubuntu):
status: New → Confirmed
Loye Young (loyeyoung) wrote :

Correcting error:

"This is so no matter which path to the swap partition I use, including: /dev/disks/by-uuid/XXXX, the device show by "grep /proc/swaps", or /dev/mapper/cryptswap1."

๐—ฆ๐—ต๐—ผ๐˜‚๐—น๐—ฑ ๐˜€๐—ฎ๐˜†:

"This is so no matter which path to the swap partition I use, including: /dev/disks/by-uuid/XXXX, the device show by "๐—ฐ๐—ฎ๐˜ /proc/swaps", or /dev/mapper/cryptswap1."

As noted by @loyeyoung, a better work-around than the one in Comment #2 may be to disable swap rather than running with an encrypted home directory and unencrypted swap as that state is a security liability. For the case where hibernate depends on swap, hibernate would also need to be disabled.

Changed in systemd (Ubuntu):
importance: Undecided → Medium
B. C. Schmerker (bcschmerker) wrote :

I have a similar issue with ubuntuยฎ 16.04.1: My own system recently hung attempting to initialize not more than two of five swap partitions ("[<---->] A start job is running for dev-mapper-cryptswap?.device (??m??s / No limit)") on several attempts and subsequently hung on unattended-upgrades on attempted reboot. Disabling cryptswap worked around the problem. The hang-on-startup problem initially appeared to be centered on hit-and-miss automatic key generation for ecryptfs; but the unattended-upgrades problem is simultaneous. Don't yet know if I've this Bug (viz., #1594035) in specific, or have to file a new Bug, as of 6 December 2016; the answer depends on replies to Question #404663, opened 5 December.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers