Skip to main content
Topic solved
This topic has been marked as solved and requires no further attention.
Topic: elogind 239.3-1 doesn't start user session (Read 2972 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

elogind 239.3-1 doesn't start user session

I just got bit by this one which can be very nasty since anything that depends/uses logind will break on you. I have yet to verify if it's an upstream issue or a packaging thing (need to do more testing), but I thought I should at least warn some people. The newest elogind did not execute properly and trying to start it with run (sudo sv start elogind) would simply timeout and the service would stay down. When running "loginctl", instead of listing a user session it said:
Code: [Select]
launch helper exited with unknown return code 127

For now, the workaround is to either just downgrade elogind in your cache or alternatively you can build elogind yourself from master. I can't imagine I'm the only one that will be hit by this, so beware.

Re: elogind 239.3-1 doesn't start user session

Reply #1
Okay I think I see the problem. It's not upstream. Building v239.3 straight from the source works. I think the built package somehow got corrupted while being pushed to the repos. I ran makepkg on the PKGBUILD and installed elogind+libelogind that way. I had no problems running the service and also noticed that both the checksums and the filesize of the package I built myself are different than what's in the system repo.

System repo's elogind's sha256sum:
Code: [Select]
cb8b3193420a3b81dc4bd11476f42d89ff91ab2f44dfed6ea64494c64e322802
(792K)

The one built from the PKGBUILD:
Code: [Select]
55c804a49b5cee435c58a7d9e10f358b01c9c83a7749976308b1301988c69602
(804K)

I believe the elogind package that is in the repo is incorrect and should be rebuilt and reuploaded.

Re: elogind 239.3-1 doesn't start user session

Reply #2
I downgrade elogind and libelogind, because Vbox error now im lost.
vbox is not working after the downgrade of elogind.

Re: elogind 239.3-1 doesn't start user session

Reply #3
Okay I think I see the problem. It's not upstream. Building v239.3 straight from the source works. I think the built package somehow got corrupted while being pushed to the repos. I ran makepkg on the PKGBUILD and installed elogind+libelogind that way. I had no problems running the service and also noticed that both the checksums and the filesize of the package I built myself are different than what's in the system repo.

System repo's elogind's sha256sum:
Code: [Select]
cb8b3193420a3b81dc4bd11476f42d89ff91ab2f44dfed6ea64494c64e322802
(792K)

The one built from the PKGBUILD:
Code: [Select]
55c804a49b5cee435c58a7d9e10f358b01c9c83a7749976308b1301988c69602
(804K)

I believe the elogind package that is in the repo is incorrect and should be rebuilt and reuploaded.

Note that sha256sum will be different per build depending on its package's .BUILDINFO file.
For example, my sha256sum of recently-built elogind-239.3-1 is 098d103c8bd8c45f13b54a7d87a6bf1c37cd507e52512e94afd2956db199024e

Also, with elogind package from system, can you see the logs from the runsvdir process if anything happened with elogind in particular? Just the ps output should be enough.
now only the dinit guy in artix

Re: elogind 239.3-1 doesn't start user session

Reply #4
Note that sha256sum will be different per build depending on its package's .BUILDINFO file.
For example, my sha256sum of recently-built elogind-239.3-1 is 098d103c8bd8c45f13b54a7d87a6bf1c37cd507e52512e94afd2956db199024e

Also, with elogind package from system, can you see the logs from the runsvdir process if anything happened with elogind in particular? Just the ps output should be enough.

Oh duh you're right, I forgot about .BUILDINFO differences. Still, I don't think the package sizes should be different though right?

I think I might be doing something wrong here (I just did runsvdir -P /etc/runit/sv/elogind), but this is what I got from runsvdir.
Code: [Select]
runsv supervise: fatal: unable to lock supervise/lock: temporary failure

ps -A doesn't show anything regarding elogind at all (which makes sense because the process doesn't even start).
Code: [Select]
  PID TTY          TIME CMD
    1 ?        00:00:00 runit
    2 ?        00:00:00 kthreadd
    3 ?        00:00:00 rcu_gp
    4 ?        00:00:00 rcu_par_gp
    5 ?        00:00:00 kworker/0:0-mm_percpu_wq
    6 ?        00:00:00 kworker/0:0H-kblockd
    8 ?        00:00:00 mm_percpu_wq
    9 ?        00:00:00 ksoftirqd/0
   10 ?        00:00:02 rcu_preempt
   11 ?        00:00:00 rcu_sched
   12 ?        00:00:00 rcu_bh
   13 ?        00:00:00 rcuc/0
   14 ?        00:00:00 rcub/0
   15 ?        00:00:00 migration/0
   16 ?        00:00:00 idle_inject/0
   18 ?        00:00:00 cpuhp/0
   19 ?        00:00:00 cpuhp/1
   20 ?        00:00:00 idle_inject/1
   21 ?        00:00:00 migration/1
   22 ?        00:00:00 rcuc/1
   23 ?        00:00:00 ksoftirqd/1
   25 ?        00:00:00 kworker/1:0H-kblockd
   26 ?        00:00:00 cpuhp/2
   27 ?        00:00:00 idle_inject/2
   28 ?        00:00:00 migration/2
   29 ?        00:00:00 rcuc/2
   30 ?        00:00:00 ksoftirqd/2
   31 ?        00:00:00 kworker/2:0-events
   32 ?        00:00:00 kworker/2:0H-kblockd
   33 ?        00:00:00 cpuhp/3
   34 ?        00:00:00 idle_inject/3
   35 ?        00:00:00 migration/3
   36 ?        00:00:00 rcuc/3
   37 ?        00:00:00 ksoftirqd/3
   38 ?        00:00:00 kworker/3:0-mm_percpu_wq
   39 ?        00:00:00 kworker/3:0H-kblockd
   40 ?        00:00:00 cpuhp/4
   41 ?        00:00:00 idle_inject/4
   42 ?        00:00:00 migration/4
   43 ?        00:00:00 rcuc/4
   44 ?        00:00:00 ksoftirqd/4
   45 ?        00:00:00 kworker/4:0-mm_percpu_wq
   46 ?        00:00:00 kworker/4:0H-kblockd
   47 ?        00:00:00 cpuhp/5
   48 ?        00:00:00 idle_inject/5
   49 ?        00:00:00 migration/5
   50 ?        00:00:00 rcuc/5
   51 ?        00:00:00 ksoftirqd/5
   52 ?        00:00:00 kworker/5:0-rcu_gp
   53 ?        00:00:00 kworker/5:0H-kblockd
   54 ?        00:00:00 cpuhp/6
   55 ?        00:00:00 idle_inject/6
   56 ?        00:00:00 migration/6
   57 ?        00:00:00 rcuc/6
   58 ?        00:00:00 ksoftirqd/6
   59 ?        00:00:00 kworker/6:0-events
   60 ?        00:00:00 kworker/6:0H-kblockd
   61 ?        00:00:00 cpuhp/7
   62 ?        00:00:00 idle_inject/7
   63 ?        00:00:00 migration/7
   64 ?        00:00:00 rcuc/7
   65 ?        00:00:00 ksoftirqd/7
   67 ?        00:00:00 kworker/7:0H-kblockd
   68 ?        00:00:00 kdevtmpfs
   69 ?        00:00:00 netns
   70 ?        00:00:00 rcu_tasks_kthre
   71 ?        00:00:00 kauditd
   72 ?        00:00:00 kworker/1:1-events
   73 ?        00:00:00 kworker/2:1-events
   74 ?        00:00:00 khungtaskd
   75 ?        00:00:00 oom_reaper
   76 ?        00:00:00 writeback
   77 ?        00:00:00 kcompactd0
   78 ?        00:00:00 ksmd
   79 ?        00:00:00 khugepaged
   80 ?        00:00:00 crypto
   81 ?        00:00:00 kintegrityd
   82 ?        00:00:00 kblockd
   84 ?        00:00:00 kworker/5:1-mm_percpu_wq
   85 ?        00:00:00 edac-poller
   86 ?        00:00:00 devfreq_wq
   87 ?        00:00:00 watchdogd
   89 ?        00:00:00 kswapd0
  132 ?        00:00:00 kthrotld
  133 ?        00:00:00 acpi_thermal_pm
  134 ?        00:00:00 nvme-wq
  135 ?        00:00:00 nvme-reset-wq
  136 ?        00:00:00 nvme-delete-wq
  137 ?        00:00:00 ipv6_addrconf
  146 ?        00:00:00 kworker/6:1-events
  147 ?        00:00:00 kstrp
  160 ?        00:00:00 charger_manager
  175 ?        00:00:00 kworker/7:1-mm_percpu_wq
  178 ?        00:00:00 nvkm-disp
  179 ?        00:00:00 ttm_swap
  180 ?        00:00:00 kworker/7:2
  184 ?        00:00:00 kworker/3:1-mm_percpu_wq
  222 ?        00:00:00 ata_sff
  235 ?        00:00:00 scsi_eh_0
  236 ?        00:00:00 scsi_tmf_0
  237 ?        00:00:00 scsi_eh_1
  238 ?        00:00:00 scsi_tmf_1
  239 ?        00:00:00 scsi_eh_2
  240 ?        00:00:00 scsi_tmf_2
  241 ?        00:00:00 scsi_eh_3
  242 ?        00:00:00 scsi_tmf_3
  243 ?        00:00:00 scsi_eh_4
  244 ?        00:00:00 scsi_tmf_4
  250 ?        00:00:00 kworker/0:2-mm_percpu_wq
  252 ?        00:00:00 kworker/2:1H-kblockd
  253 ?        00:00:00 kworker/1:1H-kblockd
  257 ?        00:00:00 kworker/5:1H-kblockd
  263 ?        00:00:00 kworker/7:1H-kblockd
  265 ?        00:00:00 kworker/4:1H-kblockd
  276 ?        00:00:00 kdmflush
  277 ?        00:00:00 kcryptd_io
  279 ?        00:00:00 kcryptd
  280 ?        00:00:02 dmcrypt_write
  282 ?        00:00:00 kworker/u17:1-kcryptd
  285 ?        00:00:00 kdmflush
  287 ?        00:00:00 kdmflush
  289 ?        00:00:00 kdmflush
  305 ?        00:00:00 kworker/6:1H-kblockd
  311 ?        00:00:00 jbd2/dm-2-8
  312 ?        00:00:00 ext4-rsv-conver
  319 ?        00:00:00 kworker/3:1H-kblockd
  321 ?        00:00:00 kworker/u17:6-kcryptd
  322 ?        00:00:00 kworker/0:1H-kblockd
  323 ?        00:00:00 kworker/u17:7-kcryptd
  444 ?        00:00:00 kworker/u16:2-events_power_efficient
  773 ?        00:00:00 udevd
  861 ?        00:00:00 irq/32-mei_me
  883 ?        00:00:00 kworker/4:4-rcu_gp
  907 ?        00:00:00 i915/signal:0
  908 ?        00:00:00 i915/signal:1
  909 ?        00:00:00 i915/signal:2
  910 ?        00:00:00 i915/signal:6
  953 ?        00:00:00 cfg80211
 1208 ?        00:00:00 jbd2/sda1-8
 1209 ?        00:00:00 ext4-rsv-conver
 1210 ?        00:00:00 jbd2/dm-3-8
 1211 ?        00:00:00 ext4-rsv-conver
 1518 ?        00:00:01 runsvdir
 1523 ?        00:00:00 runsv
 1524 ?        00:00:00 runsv
 1525 ?        00:00:04 runsv
 1526 ?        00:00:00 runsv
 1527 ?        00:00:00 runsv
 1528 ?        00:00:00 runsv
 1529 ?        00:00:00 runsv
 1530 ?        00:00:00 runsv
 1531 tty5     00:00:00 agetty
 1532 ?        00:00:00 runsv
 1533 ?        00:00:00 runsv
 1534 tty6     00:00:00 agetty
 1535 ?        00:00:00 runsv
 1536 ?        00:00:00 runsv
 1537 ?        00:00:00 runsv
 1538 tty4     00:00:00 agetty
 1539 tty2     00:00:00 agetty
 1540 ?        00:00:00 login
 1543 ?        00:00:00 cupsd
 1544 ?        00:00:04 dbus-daemon
 1545 ?        00:00:00 syslog-ng
 1546 tty3     00:00:00 agetty
 1551 ?        00:00:03 mysqld
 1552 ?        00:00:00 dhcpcd
 1700 tty1     00:00:00 zsh
 1713 ?        00:00:00 dbus-daemon
 1751 ?        00:00:00 gvfsd
 1774 ?        00:00:00 gvfsd-fuse
 1809 ?        00:00:35 pulseaudio
 1826 ?        00:00:00 at-spi-bus-laun
 1831 ?        00:00:00 dbus-daemon
 1833 ?        00:00:00 at-spi2-registr
 1879 ?        00:00:00 gsettings-helpe
 2370 ?        00:00:00 kworker/u17:5-kcryptd
 3610 ?        00:00:00 kworker/2:2
 3805 ?        00:00:00 dbus-daemon
 3932 tty1     00:00:00 ps
10621 ?        00:00:00 kworker/1:0-events
10633 ?        00:00:00 scsi_eh_5
10634 ?        00:00:00 scsi_tmf_5
10635 ?        00:00:10 usb-storage
10655 ?        00:00:00 gvfs-udisks2-vo
10659 ?        00:00:00 udisksd
10664 ?        00:00:02 polkitd
10700 ?        00:00:00 gvfsd-trash
10741 ?        00:00:00 gvfsd-metadata
11022 ?        00:00:00 gconfd-2
12330 ?        00:00:01 kworker/u16:4-events_unbound
19983 ?        00:00:00 kworker/u17:0-kcryptd
22729 ?        00:00:00 kworker/u16:3-flush-254:3
22905 ?        00:00:00 kworker/u17:4-kcryptd
23284 ?        00:00:01 runsv
28445 ?        00:00:00 kworker/u16:0-events_power_efficient
29380 ?        00:00:00 kworker/u17:3-kcryptd
30236 ?        00:00:00 kworker/u17:8-kcryptd
32213 ?        00:00:00 kworker/u17:2-kcryptd

Re: elogind 239.3-1 doesn't start user session

Reply #5
ps -A doesn't show anything regarding elogind at all (which makes sense because the process doesn't even start).

Just do ps aux | grep runsvdir. The COMMAND field should give you something.

And no. runsvdir -P /etc/runit/sv/elogind won't work since it will only work for full services directory (e.g. /run/runit/service)
now only the dinit guy in artix

Re: elogind 239.3-1 doesn't start user session

Reply #6
Just do ps aux | grep runsvdir. The COMMAND field should give you something.

And no. runsvdir -P /etc/runit/sv/elogind won't work since it will only work for full services directory (e.g. /run/runit/service)

Ah that seems like the right approach indeed. Here's what I get from the elogind package from the repo.
Code: [Select]
root      1402  0.0  0.0   2328  1168 ?        Ss   18:25   0:00 runsvdir -P /run/runit/service log: /elogind: error while loading shared libraries: libaudit.so.1: cannot open shared object file: No such file or directory /usr/lib/elogind/elogind: error while loading shared libraries: libaudit.so.1: cannot open shared object file: No such file or directory /usr/lib/elogind/elogind: error while loading shared libraries: libaudit.so.1: cannot open shared object file: No such file or directory 
dudemanguy      1844  0.0  0.0   9220  2412 tty1     S+   18:26   0:00 grep runsvdir

And here's what I get from the one I built myself from the PKGBUILD
Code: [Select]
dudemanguy       885  0.1  0.1  66484 14164 pts/2    Sl+  21:17   0:00 nvim runsvdir-output
dudemanguy      1207  0.0  0.0   9220  2400 pts/3    S+   21:18   0:00 grep runsvdir
root      1402  0.0  0.0   2328  1168 ?        Ss   18:25   0:00 runsvdir -P /run/runit/service log: sing servicehelper) dbus-daemon[1425]: [system] Successfully activated service 'org.freedesktop.UPower' .udisks-Message: 18:40:53.728: Mounted /dev/sdd1 at /run/media/dudemanguy/62AD-3FBB on behalf of uid 1000 udisks-Message: 18:54:59.782: Cleaning up mount point /run/media/dudemanguy/62AD-3FBB (device 8:49 is not mounted) udisks-Message: 18:54:59.819: Unmounted /dev/sdd1 on behalf of uid 1000 ..........

Re: elogind 239.3-1 doesn't start user session

Reply #7
Or install "audit"  because a dependency for libaudit.so.1 has crept in there, perhaps picked up from the build environment somehow. Nice to find audit available as a package anyway  ;D

Code: [Select]
$ readelf -a /usr/lib/elogind/elogind |ag library
 0x0000000000000001 (NEEDED)             Shared library: [libpthread.so.0]
 0x0000000000000001 (NEEDED)             Shared library: [libc.so.6]
 0x0000000000000001 (NEEDED)             Shared library: [libelogind-shared-239.3.so]
 0x0000000000000001 (NEEDED)             Shared library: [libacl.so.1]
 0x0000000000000001 (NEEDED)             Shared library: [libudev.so.1]
 0x0000000000000001 (NEEDED)             Shared library: [libaudit.so.1]
 0x0000000000000001 (NEEDED)             Shared library: [ld-linux-x86-64.so.2]
 0x000000000000000f (RPATH)              Library rpath: [/usr/lib/elogind]

Re: elogind 239.3-1 doesn't start user session

Reply #8
Got bit by this one just now after system update. On reboot, I encountered error during open-rc starting services stating...
Code: [Select]
* Starting elogind ...
* start-stop-daemon: failed to start `//usr/lib/elogind/elogind'
* ERROR: elogind failed to start

A quick fix to /etc/conf.d/elogind to remove the extra slash and I tried starting the service again. Still a fail.

I attempted to start elogind manually from the command line...
Code: [Select]
# /usr/lib/elogind/elogind -D
and was greeted with an error message stating that libaudit.so.1 was missing.

A quick pacman search lead me to...
Code: [Select]
$ pacman -Qi audit
Name            : audit
Version         : 2.8.4-2
Description     : Userspace components of the audit framework
Architecture    : x86_64
URL             : https://people.redhat.com/sgrubb/audit
Licenses        : GPL
Groups          : None
Provides        : libaudit.so=1-64  libauparse.so=0-64
Depends On      : krb5  libcap-ng
Optional Deps   : None
Required By     : None
Optional For    : None
Conflicts With  : None
Replaces        : None
Installed Size  : 1048.00 KiB
Packager        : Artix Build Bot <[email protected]>
Build Date      : Wed 03 Oct 2018 02:05:06 AM MDT
Install Date    : Thu 13 Dec 2018 04:02:17 AM MST
Install Reason  : Explicitly installed
Install Script  : No
Validated By    : Signature
...seems there is an undeclared dependency, as some else pointed out above.


Installed system/audit. Tried starting elogind service again... and success.
The elogind package has upstream problems.

Re: elogind 239.3-1 doesn't start user session

Reply #9
But here's the thing, I don't need audit if I build it from source myself. Directly from git master, the V239 branch, or even the PKGBUILD from the gitea repo. I'm using 239.3 without audit installed right now. The mysterious audit dependency only occurs if I use the package from the repo and I have no idea why. It's almost as if the elogind package in the repo was built using a different PKGBUILD than the one in the gitea repo.

I don't think there's anything wrong upstream. I think something went weird with the packaging.

Re: elogind 239.3-1 doesn't start user session

Reply #10
I moved elogind too early in system while dbus is still in gremlins, which has the audit depend.

Anyway, please note, that if you use makepkg on your system, a lot of configure or meson builds autoset some feature, ie makepkg does not produce deployable builds for repos. It always depends what is installed on your system if you use makepkg.

 

Re: elogind 239.3-1 doesn't start user session

Reply #11
Anyway, please note, that if you use makepkg on your system, a lot of configure or meson builds autoset some feature, ie makepkg does not produce deployable builds for repos. It always depends what is installed on your system if you use makepkg.


Can you explain that a bit more for me ?

I was under the impression, that so long as a PKGBUILDS's 'requires=' is complete and correct then the package should work on any system with the correct architecture . I know that if you have very specific processor flags set in makepkg.conf then the resulting binaries may not work on other machines with a different processor / extensions.  But where generic flags are used, x32 or x64, then the package should work on any machine that is either x32 or x64 as applicable

I thought the "configure or meson builds autoset some feature" was, or should be, controlled by the PKGBUILD ?

When I used to maintain some PKGBUILDS's for arm developer boards (Odroid mainly) I used to build in a clean Chroot (https://wiki.archlinux.org/index.php/DeveloperWiki:Building_in_a_clean_chroot) to make sure I was not missing dependencies)

Sorry to go a bit off topic, and not saying you are wrong. Just that I don't understand :)

Re: elogind 239.3-1 doesn't start user session

Reply #12

When I used to maintain some PKGBUILDS's for arm developer boards (Odroid mainly) I used to build in a clean Chroot (https://wiki.archlinux.org/index.php/DeveloperWiki:Building_in_a_clean_chroot) to make sure I was not missing dependencies)



Artix packages are all built in chroot.

Suppose some software autidetects a library during build.
With makepkg, this library/feature would eventually get enabled, if you have this library installed.
With chroot, if this library is not in the depends or makedepends, it would get disabled.
Chroot is isolated build environment, while normal user system is not, it depends what the user has installed

Re: elogind 239.3-1 doesn't start user session

Reply #13
Chroot is isolated build environment, while normal user system is not, it depends what the user has installed

Cheers  :)
I realise now I'd half answered my own question.
Crystal clear again.
I'm just hard of thinking sometimes ! 

Re: elogind 239.3-1 doesn't start user session

Reply #14
And now, all is fixed with 293.3-3. Awesome.