FS#44016 - [systemd] journald regression in v219 makes boot hang

Attached to Project: Arch Linux
Opened by Christian Hesse (eworm) - Monday, 02 March 2015, 14:46 GMT
Last edited by Dave Reisner (falconindy) - Thursday, 30 July 2015, 05:50 GMT
Task Type Bug Report
Category Packages: Testing
Status Closed
Assigned To Thomas Bächler (brain0)
Dave Reisner (falconindy)
Tom Gundersen (tomegun)
Architecture All
Severity Critical
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 16
Private No

Details

Description:
A journald regression in v219 makes the boot hang. This has been reported for Debian [0] and Ubuntu [1] and a discussion has been started on systemd mailing list [2].
I've seen this with Arch package system 219-2 as well.

Looks like the be fix for now is to revert commit add4bf64 ("journald: allow restarting journald without losing stream connections") [3].

[0] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=778970
[1] https://bugs.launchpad.net/bugs/1423811
[2] http://comments.gmane.org/gmane.comp.sysutils.systemd.devel/28525
[3] http://cgit.freedesktop.org/systemd/systemd/commit/?id=13790add4bf64

Additional info:
systemd 219-2
libsystemd 219-2

Steps to reproduce:
Update to system 219-2, then reboot again and again until boot hangs.
This task depends upon

Closed by  Dave Reisner (falconindy)
Thursday, 30 July 2015, 05:50 GMT
Reason for closing:  Fixed
Additional comments about closing:  systemd-220-1
Comment by Dave Reisner (falconindy) - Monday, 02 March 2015, 14:52 GMT
Yep, this is what was keeping the package in testing. Thanks for the link.
Comment by Dave Reisner (falconindy) - Saturday, 14 March 2015, 15:22 GMT
To be clear, I'm not going to be reverting this patch. There needs to be traction upstream to get this properly fixed.
Comment by Joseph Riches (joebonrichie) - Sunday, 15 March 2015, 15:27 GMT
Seemed to have this issue on systemd-219-3 on Acer C720
Just downgraded to systemd-219-2 and could boot fine again.

edit: disregard this seems to be something unrelated
Comment by taylorchu (taylorchu) - Monday, 06 April 2015, 00:57 GMT
Running systemd 219-4. And this issue bites me almost every reboot.
Comment by Dave Reisner (falconindy) - Monday, 06 April 2015, 01:06 GMT
I can do without the empty +1-style posts. Work with upstream if you want this fixed. The situation has not changed in git.
Comment by Christian Hesse (eworm) - Sunday, 19 April 2015, 20:07 GMT
Now that systemd 219-5 is in [core]... I do not see any changes related to the issue, no?
Comment by Dave Reisner (falconindy) - Sunday, 19 April 2015, 20:23 GMT
No changes here because there weren't any upstream. I opted to not block the the gnutls rebuild and let this move to [core]. If you're affected by this, then help upstream resolve the bug.
Comment by Christian Hesse (eworm) - Wednesday, 29 April 2015, 10:03 GMT
Good news! Looks like this is fixed in git master:

http://lists.freedesktop.org/archives/systemd-devel/2015-April/031348.html
http://cgit.freedesktop.org/systemd/systemd/commit/?id=64144440

I did not verify myself, though.
Comment by Christian Hesse (eworm) - Wednesday, 29 April 2015, 11:24 GMT
It's not true that this applies cleanly to v219... We need this in addition:
http://cgit.freedesktop.org/systemd/systemd/commit/?id=d4a144fa

Just compiling...
Comment by Christian Hesse (eworm) - Wednesday, 29 April 2015, 11:48 GMT
As this does happen only very seldom... Anybody wants to give it a try?
The more testers we have the better test results we have.
   systemd.diff (1.7 KiB)
Comment by Benjamin Robin (benjarobin) - Sunday, 03 May 2015, 10:31 GMT
I tried the patched PKGBUILD on 2 computers which suffer this problem, and there is no longer any boot hang.
Comment by Christian Hesse (eworm) - Monday, 04 May 2015, 07:56 GMT
Using this since Wednesday without issues.
Comment by Eivind (mokkurkalve) - Monday, 04 May 2015, 14:02 GMT
This thing hit me bad. The patch fix it good....
Comment by Francis Herne (FLHerne) - Tuesday, 19 May 2015, 19:07 GMT
This is now fixed in upstream git (by accident, per later emails in linked thread! :-/) http://lists.freedesktop.org/archives/systemd-devel/2015-April/031348.html
Hopefully it'll be backported soon?
Comment by Christian Hesse (eworm) - Tuesday, 19 May 2015, 19:12 GMT
That is what I referenced about three weeks ago.
Feel free to grab systemd.diff from above and build a fixed package.
Comment by Christian Hesse (eworm) - Monday, 25 May 2015, 18:26 GMT
This should be fixed with systemd 220-1 in [testing].
Comment by JF (jfcandidofilho) - Friday, 05 June 2015, 17:39 GMT
Well.. It took me time to arrive here since I'm less than a n00b. But I wanted to report it the day before yesterday because it's when I updated the system and everything went bananas.

Going to the point..: yesterday, after some hours of coding - by luck I save every microsecond - the X went down the hill towards the cliff. I tried to start the X by typing "startx" all the time, even after some reboots, but nothing: colord.service at fault (I guess..?) because of systemd (another guess).

After looking, I noticed a warning at boot time. "colord.service" couldn't start. And a little searching (googling) pinpointed that, time to time, systemd has this strange behaviour of not starting things (ie. bug).

So here I am to report that some update in these last days (even, maybe, prior to this month that just tarted) broke things. But I can't confirm what the problem is - I think don't have the knowledge.

There are some logs attached (I don't know if they suffice) but I just wanted to say that I am using the testing repository to test if with systemd 220.3 things works (the problematic one is 219). And till know, they do work. I hope 220 is a real fix and that this update comes as soon as possible.

Files:
:: journalctl -p 0..3 -xn > 1.txt
:: systemctl status colord.service > 2.txt
:: journalctl -xe > 3.txt

EDIT - 2015-06-16: system crashed, again. This new version doesn't work OK. I lasted 11 days. I got the same files and made a diff but relevant info hasn't changed. So, 220 IS NOT a fix.
   1.txt (2.5 KiB)
   2.txt (0.9 KiB)
   3.txt (120.8 KiB)

Loading...