regressions: hibernation issues from 254.5 to 255~rc2

[calestyo]

systemd version the issue has been seen with

255~rc2

Used distribution

Debian sid

Linux kernel version used

6.5.0-4-amd64

CPU architectures issue was seen on

x86_64

Component

systemctl, systemd, systemd-logind

Expected behaviour you didn't see

Hey.

Forwarding this from Debian bug #1056135.

Since upgrading from 254.5 to 255~rc2, I see the following two issues (which both disappear when downgrading to 254.5-1 (and completely rebooting the system, before trying again - they also shop up again, only after completely rebooting the upgraded systemd again)):

1. I have a somewhat special hibernation setup.
   Hibernation goes to a btrfs swap file on top of dm-crypt.
   The following unit files are used for that:
   /etc/systemd/system/data-swap-hibernation.swap:

    [Unit]
    Before=systemd-hibernate.service
    StopWhenUnneeded=true
   
    [Swap]
    What=/data/swap/hibernation
    Options=noauto
   
    [Install]
    RequiredBy=systemd-hibernate.service
   

   I also have a /etc/systemd/system/hibernation-cleanup.service:

    [Unit]
    Description=synchronise cached writes to persistent storage and free page cache before hibernation in order to minimise the image
    Before=hibernate.target
   
    [Service]
    Type=oneshot
    ExecStart=sync
    ExecStart=@sh %n:sh -c 'printf 1 >/proc/sys/vm/drop_caches'
   
    [Install]
    WantedBy=hibernate.target
   

   but I don't think this has anything to do with the issue here.
   
   There are symlinks:

    /etc/systemd/system/systemd-hibernate.service.requires/data-swap-hibernation.swap -> /etc/systemd/system/data-swap-hibernation.swap
    /etc/systemd/system/hibernate.target.wants/hibernation-cleanup.service -> /etc/systemd/system/hibernation-cleanup.service
   

   Especially, the hibernation file may not be availalbe (or activated as swap), yet, when I go into hibernation.
   This caused desktop environments (I use Cinnamon) not to show the "Hibernate" button, which I've been able to solve via a:
   /etc/systemd/system/systemd-logind.service.d/disable-hibernation-swap-check.conf:

    [Service]
    Environment="SYSTEMD_BYPASS_HIBERNATION_MEMORY_CHECK=1"
   

   However, starting with 255~rc2-1 (but solved when going back to 254.5-1) this no longer seems to have an effect. At least Cinnamon then doesn't show the Hibernate button anymore.

2. When, because of (1), manually calling:

   # systemctl hibernate
   Call to Hibernate failed: No such file or directory
   

   The above error occurs (which itdoes not with 254.5-1), but the system then still seems to hibernate normally (and can also resume from that).

Notice also that I've changed logind.conf:

[Login]
HandleSuspendKey=ignore
HandleSuspendKeyLongPress=ignore
HandleLidSwitch=lock
HandleLidSwitchExternalPower=lock
HandleLidSwitchDocked=lock

and sleep.conf:

[Sleep]
HibernateMode=shutdown platform

Any ideas?

Thanks,
Chris.

Unexpected behaviour you saw

No response

Steps to reproduce the problem

No response

Additional program output to the terminal or log subsystem illustrating the issue

No response

[calestyo]

Some logs, which I've made with:

[Manager]
LogLevel=debug

in /etc/systemd/system.conf and after a fresh restart of the system with that.

1. journalctl -b output, but only the part around the time where within Cinnamon I pressed the Power-Button, which causes the dialogue that asks for Shutdown/Restart/Hibernate/Suspend, etc. to show up, without actually triggering any of these:
   systemd.debug.powerbutton-handler.txt
2. journalctl -b output, but only the part around executing systemctl hibernate until having the system resumed again:
   systemd.debug.hibernate+resume.txt
3. journalctl -b | grep -E 'cinnamon|login':
   systemd.debug.cinnamon+login.txt

If you need anything else, please tell me what :-) (I've been a bit reluctant to simply send the whole journal since it's so much any I didn't want to check all for any sensitive information ^^)

Thanks,
Chris.

[yuwata]

cc @YHNdnzj ?

[YHNdnzj]

Indeed. Such setups are unintentionally invalidated in #29382, where more safety checks for hibernation were added. #30074 should help to generate more helpful error message.

TBH, I'm not sure if this kind of setup is worth supporting. I mean, the swap space is always around, so why not keep it enabled? See more justifications of mine in https://wiki.archlinux.org/title/Talk:Power_management#On-demand_swap_for_hibernation_only . But, since it breaks compatibility, I'll submit a fix.

[YHNdnzj]

@calestyo Could you please try #30074?

[calestyo]

But I shall still recommend https://chrisdown.name/2018/01/02/in-defence-of-swap.html to everyone ;-)

Well, I've been in the discussion about swap being considered nowadays again a good thing before, and I do understand the point that one cannot page anonymous memory without swap... but still, I think there are some niche cases, where it makes sense to run without general swap, but were one might still want it for hibernation.

So it's really good, that you keep supporting this.

I've tested #30074, and when I rebooted the system with a build of systemd that had the patches applied, it froze at boot (while systemd was already doing its thing) and CPU went to 100% (assuming from the fan noise).

Had never seen this on that system before. Only a forcible poweroff helped. Trying to boot again with no further changes, worked like a charm.

Also, the commits fixed both issues. Especially systemdct hibernate now again causes the screensaver to lock up the DE.

Thanks for the fix :-)