elogind-241.4

[Douglas R. Reno]

New minor version

Note that the currency script does not pick it up.

My sed will still be required, so I'll pick this up and test it.

[Douglas R. Reno]

This is the fourth release of the elogind-241 series, updated to systemd-stable commit 323cdf4d4d.

Warning: It was reported that waking up from hibernation results in a kernel crash on at least one machine when run with a 5.3.x kernel. Kernels <5.3.0 are not affected.

Fixes:

• cgroup-util: kill also live threads from zombie processe.
• fix build with compilers with default stack-protector enabled.
• login/elogind-dbus.c: Wakeup sleeping processes also when suspend/hibernate failed.

Changes / Additions:

• Upgrade README to README.md and add Travis CI badges for quick status display
• sd-bus: when running user find function don't trust the value to be initialized
• shared/conf-parser: be nice and ignore lines without "="

---

I will give hibernation a shot too, and potentially put in something to modify the config in the book if it causes a panic.

[Pierre Labastie]

What about this sed (it's a little tricky, because there are several places with "read_one_line_file"

sed 's|r = read_one.*(p,|r = parse_env_file(NULL, p, "UID",|' -i src/basic/cgroup-util.c

[Bruce Dubbs]

Here is what I suggest:

sed -e '/dirent-util/a #include "env-file.h"' \
    -e '/sessions/{n;n;s/=.*$/=  parse_env_file(NULL, p, "UID", \&s);/}' \
    -i src/basic/cgroup-util.c

[Pierre Labastie]

Ah yes, hadn't seen the added header, sorry... For the other line, both sed's are fragile if something changes. Mine relies on spaces being there, but Bruce's relies on a blank line being there... But hopefully, next time this file is edited will be for fixing this issue.

[Bruce Dubbs]

Replying to pierre.labastie:

Ah yes, hadn't seen the added header, sorry... For the other line, both sed's are fragile if something changes. Mine relies on spaces being there, but Bruce's relies on a blank line being there... But hopefully, next time this file is edited will be for fixing this issue.

You're right, it does depend on white space. When I do package updates, I generally look at all seds and patches for the package to see if they are still needed and, if so, still valid.

[Pierre Labastie]

Just tried the sed (either one) with 241.3, and gdm is OK now!

[Douglas R. Reno]

That's good. I'm working on the stack, but have enabled hibernation on my system so I can see if that hang still happens. I just encountered a dependency on gstreamer, so it's time for that to get built. If the crash happens, I'll troubleshoot it.

[Pierre Labastie]

With 241.3, no crash. Actually, there are no options in gnome for suspend/hibernation, but you can associate hibernation to the power button (in settings->Power). I've tried that, and it worked OK. Is the crash suppoed to be specific to 241.4?

[Douglas R. Reno]

Yes it is, and it can happen in other desktop environments as well from what I gather by looking at it. It happens when the kernel brings the system back up

[Pierre Labastie]

OK, Will give it a try (on gnome only, for now).

[Pierre Labastie]

Hibernate worked with elogind 241.4, and suspend too. Will build and try lxde.

[Douglas R. Reno]

Alright cool, would you like to do the update? I can still do it, but looking at the end of the day or early tomorrow before I will get around to it.

[Pierre Labastie]

Hmmm, resuming after hibernate on lxde did not work very well. I had to pass the mouse pointer everywhere on the screen to have it redrawn... Will try to downgrade to 241.3

[Pierre Labastie]

And the keyboard layout is "us" (I think).

[Pierre Labastie]

Ah, this comes from starting lxde from gdm (the keyboard problem, I mean).

[Douglas R. Reno]

I can probably look into that tomorrow. Stepping away for now, my left eye is twitching and it's making it very difficult to focus on my screen.

[Pierre Labastie]

And here, it is getting late. Let's retry tomorrow...

[Pierre Labastie]

The hibernate problem on lxde is the same with 241.3. But I have to test (tomorrow) with starting Xorg from a terminal.

[Douglas R. Reno]

Now that my technical difficulties with my workstation are resolved, I'll continue on this in the morning - it's 23:09 here

[Pierre Labastie]

Replying to pierre.labastie:

Ah, this comes from starting lxde from gdm (the keyboard problem, I mean).

No: this comes from /etc/X11/xorg.conf.d/30-keyboard.conf being owned by root and only readable by user. This is a bug in blocaled, will report upstream :)

Concerning hibernate, it has exactly the same problem when starting from command line... But since it is not specific to 241.4, I guess we can update (and possibly make another ticket for following up the hibernate issue).

[Douglas R. Reno]

Sounds good to me. Continuing work on that system this morning

[Pierre Labastie]

Doug, I am ready to commit the update, may I?

[Douglas R. Reno]

Please do so

[Pierre Labastie]

Fixed at r22487

[Douglas R. Reno]

I just restarted the elogind service on my system after upgrading to 241.4, and had a problem where the X server crashed (I expected that would happen), but keyboard and mouse control was not returned to my TTY. I was unable to use Ctrl+Alt+Del to reboot either, I had to go back to my workstation and restart it over SSH.

Has anyone else experienced this? My concern is more with the lack of keyboard input on the console. No TTY changes are possible either

[Pierre Labastie]

Replying to renodr:

I just restarted the elogind service on my system after upgrading to 241.4, and had a problem where the X server crashed (I expected that would happen), but keyboard and mouse control was not returned to my TTY. I was unable to use Ctrl+Alt+Del to reboot either, I had to go back to my workstation and restart it over SSH.

Has anyone else experienced this? My concern is more with the lack of keyboard input on the console. No TTY changes are possible either

Do you mean you just restarted elogind without rebooting? I've never done that. But I would expect all the seats and sessions to be lost, so experiencing permission problems everywhere (and I guess that might prevent keyboard and mouse input).

But what I have seen is that alt-ctrl-fx does not work when you start from gdm. Or it does (for example, you can type alt-ctrl-f2) but then you get a black screen without a login prompt.

[Douglas R. Reno]

Yes I restarted elogind without rebooting... for most other system services, I can get away with that :-) I'll write a sticky note and put it on top of that system because otherwise I know I'll forget that. No keyboard input is accepted at a TTY - it acts like the keyboard itself was disabled.

That GDM issue is weird. It works on systemd, so I don't know why it wouldn't work in elogind...

[Douglas R. Reno]

On second thought, this might be my configuration. I still have the bootscript installed.