Am Freitag, 4. Januar 2019, 12:22:46 CET schrieb Michael Biebl:
> Am 04.01.19 um 11:48 schrieb Patrick Häcker:
> > Hello,
> > 
> >> Can you post the output of
> >> ls -ld /var/lib/systemd/timesync
> >> ls -la /var/lib/private/systemd/
> > 
> > root@mmm /h/pat# ls -ld /var/lib/systemd/timesync
> > lrwxrwxrwx 1 root root 27 Jan  3  2018 /var/lib/systemd/timesync -> ../
> > private/systemd/timesync/
> > 
> > root@mmm /h/pat# ls -la /var/lib/private/systemd/
> > insgesamt 12
> > drwxr-xr-x 3 root             root             4096 Jan  3  2018 ./
> > drwx------ 3 root             root             4096 Jan  3  2018 ../
> > drwxr-xr-x 2 systemd-timesync systemd-timesync 4096 Okt 19  2017 timesync/
> 
> Hm, but this directory is writable by systemd-timesync, so
> systemd-timesyncd should have no problem with updating the timestamp.

Yes, /var/lib/private/systemd is writable. However, /var/lib/private is not 
(note the ".." entry in the listing above). And as permission is denied on 
that directory level, the subdirectory systemd can't be accessed neither.
With DynamicUser systemd would use the trick to have the separate mount 
namespace with the extended permissions due to a tmpfs on /var/lib/private, 
but as that's not active it does not save the day.

> For completeness sake, can you also post the output of
> stat /var/lib/systemd/timesync/clock
> Make that
> stat /var/lib/private/systemd/timesync/clock

root@mmm /h/pat# env LANG=EN stat /var/lib/systemd/timesync/clock
  File: /var/lib/systemd/timesync/clock
  Size: 0               Blocks: 0          IO Block: 4096   regular empty file
Device: 801h/2049d      Inode: 789553      Links: 1
Access: (0644/-rw-r--r--)  Uid: (  111/systemd-timesync)   Gid: (  135/
systemd-timesync)
Access: 2019-01-01 11:13:04.934514106 +0100
Modify: 2019-01-01 11:13:04.934514106 +0100
Change: 2019-01-01 11:13:04.934514106 +0100
 Birth: -

> (I missed that you already deleted /var/lib/systemd/timesync/)
Please note, that I somewhat mix the output from the two affected systems here 
due to system downtime of one system. This is quite practical here, as I 
haven't deleted the symbolic link on the system with the output above. So UID, 
GID and timestamps are a bit inconsistent along my entries to this bug report, 
but the rest should basically be identical.

I think the state to see the problem is quite reproducible: First install 
systemd 239 and afterwards install systemd 240, i.e. I can recreate whatever 
state you need.

Kind regards and thanks
Patrick

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply via email to