| Summary: | Shutdown and reboot buttons in KDE 4.8.0 result in a log out, back to KDM (with systemd) | ||
|---|---|---|---|
| Product: | Mageia | Reporter: | Rémi Verschelde <rverschelde> |
| Component: | RPM Packages | Assignee: | Colin Guthrie <mageia> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | Normal | CC: | AndrewL733, benbullard79, dadrown1, fri, jarillon, juan.baptiste, kernelgremlin, laidlaws, lmenut, mageia, mageia, magnus.mud, mail, mriganka13484+floss, ottosa, smiling.diego |
| Version: | Cauldron | ||
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| URL: | http://bugs.archlinux.org/task/28386 | ||
| See Also: | http://bugs.kde.org/show_bug.cgi?id=293978 | ||
| Whiteboard: | |||
| Source RPM: | systemd | CVE: | |
| Status comment: | |||
| Bug Depends on: | |||
| Bug Blocks: | 2120 | ||
|
Description
Rémi Verschelde
2012-02-20 15:40:22 CET
Magnus Rasche
2012-02-20 15:42:16 CET
CC:
(none) =>
magnus.mud same prob with reboot from kde-menue I confirm this issue still applies for Beta1 (I had only tested it on my Cauldron workstation, regularly updated since the release of Mageia 1). I have this problem too. Bug present after fresh install with beta1 64bit. CC:
(none) =>
kernelgremlin This bug happens only with systemd. No problem with sysvinit. c.f. https://bugs.archlinux.org/task/28386 https://bugs.kde.org/show_bug.cgi?id=293978 URL:
(none) =>
http://bugs.archlinux.org/task/28386
diego w
2012-02-23 20:30:43 CET
CC:
(none) =>
smiling.diego Me too. IBM Thinkpad T40. cauldron current. Interestingly: performing the same in LXDE it briefly trows up a dialog to enter root password it say is needed for this. I have not managed to be quick enough to type it, but managed to press some info button and read something about ...dbus...session... i think. May be related: On this system Hibernate and suspend are also broken, and it have problem shutting down https://bugs.mageia.org/show_bug.cgi?id=4475 CC:
(none) =>
fri The archlinux bug in #6 above is pretty active and have a couple workarounds. I tried session managers GDM, and XDM: both there is not this problem. NOTE: reboot needed for it to work properly. - it also made my suspend work :) on RazorQt it works as expected, on the same system KDE logs out I also experience bug #4475 as Morgan does. something has changed... on my setup (with auto login) it still logs out to KDM, but... if I'm logging back in to KDE4 then the shotoff works. now rebooting again to verify something about 4559 I did not experience this bug today, with an up-to-date Cauldron system using systemd. Shutdown was straightforward and didn't bring me back to KDM. I'll check a second time to be sure. @Rémi, are you using autologin? I'll update again and retest, there are quite a few 4.8.0-2 RPMs in the queue... still same: shuting down when logged in manually, just logging out when auto logged in. Unfortunately, the bug is still valid. I did another test and I was sent to KDM. I use autologin.
Juan Luis Baptiste
2012-02-28 15:08:56 CET
CC:
(none) =>
juan.baptiste i too have this problem M2B1 with all current updates i dont know if this will help you track down the problem but my work around for this is to: CTRL-ESC then i kill systemd-logind after that when i shutdown i do not go back to the login screen CC:
(none) =>
dadrown1 I simply pull down the Shutdown menu in KDM. It is quicker. (In reply to comment #20) > I simply pull down the Shutdown menu in KDM. It is quicker. i dont get that far ,as soon as i boot up i kill systemd-logind, then do updates,some testing,then when i shutdown it shuts down and does not go to kdm login I have tested this in xfce, which uses GDM. There is no problem there; it is confined to KDM. I am just downloading another 330 packages of KDE. They seem to do this every few days, with a hope and a prayer that this time it will work. It never does. (In reply to comment #25) > I am just downloading another 330 packages of KDE. They seem to do this every > few days, with a hope and a prayer that this time it will work. It never does. Unless someone actually works on the problem, then it's unlikely to make a difference. Also, such comments on a bug report are not overly useful. This is just one problem among many. Try not to get too upset about it - engineering resources are sadly not infinite! For reference, as mentioned on the upstream or arch bug report linked above, it's a combinatorial thing. KDM+KDE is the problem. If you use GDM+KDE it's fine, if you use LXDM+KDE it's fine if you use KDM+GNOME it's fine... the broken combo is KDM+KDE. (In reply to comment #26) [...] the > broken combo is KDM+KDE. more exactly, systemd+KDM+KDE :-) and the "culprit" is ... 'logind: if we have to stop a session, kill at least its leader' http://cgit.freedesktop.org/systemd/systemd/commit/?id=9b221b63e5cc62439b32bb487775856a78c6015a since this commit, systemd-logind blindly kills the leader process. if I delete the 3 lines that kill leader process, shutdown and reboot works fine. @@ -626,11 +626,6 @@ * session... */ r = manager_get_session_by_pid(s->manager, s->leader, &t); - if (r > 0 && t == s) { - kill(s->leader, SIGTERM); /* for normal processes */ - kill(s->leader, SIGHUP); /* for shells */ - kill(s->leader, SIGCONT); /* in case they are stopped */ - } } r = cg_is_empty_recursive(SYSTEMD_CGROUP_CONTROLLER, s->cgroup_path, true); I think it's really not safe to kill processes like this is done by systemd-logind, the login manager will exit itself. Assignee:
balcaen.john =>
mageia Sadly the diff you showed there cuts off the comment.
/* We still send a HUP to the leader process,
* even if we are not supposed to kill the
* whole cgroup. But let's first check the
* leader still exists and belongs to our
* session... */
So why is KDM belonging to the user session? It certainly shoudn't. And even in this case why is it different from a KDM->GNOME vs. a KDM->KDE sessions. Smells like KDM is doing strange things when logging into a KDE sessions - perhaps to try and get a smooth transition and thus the KDM process itself is somehow escaping it's cgroup... who knows. Needs investigation.
I suspect strongly that KDM is doing silly things here and while it might have worked, it's likely very bad practice and now causes breakage.
IMHO it's now working again on 4.8.1. But not 100% sure, needs more testers to confirm. CC:
(none) =>
sander.lepik The bug is still there, beta1 just updated. What can I do? If I read Colin's post correctly, it is really up to KDE to fix it. I have just booted out of Cauldron with all updates, and it was present then. It isn't just Mageia with the problem. archlinux has reported it as well. I could say: "Didn't I say so?" but as Colin says, that won't progress the bug.
Ben Bullard
2012-03-12 23:51:19 CET
CC:
(none) =>
benbullard79 Should be fixed in latest kdebase4-workspace thanks to patch from Rex Dieter @ Fedora. Status:
NEW =>
RESOLVED confirmed thanks I can confirm too that this has been resolved. Thanks. Hi Colin, This recent commit in systemd "logind: close FIFO before ending sessions cleanly", http://cgit.freedesktop.org/systemd/systemd/commit/?id=75c8e3cffd7da8eede614cf61384957af2c82a29 fixes this bug with KDE/KDM. I tested our current systemd patched with the logind FIFO patch + our kdebase4-workspace without the fedora patch, and shutdown and reboot work fine. As the problem is now fixed in systemd, the fedora's patch will probably never reviewed and pushed upstream in KDE. So, if it's possible, I would prefer that we use the systemd official patch and drop the fedora workaround. WDYT? regards, Luc @Luc, yup confirmed here also. systemd updated and kdebase4-workspace also updated and submitted for build. Thanks for letting me know - I'll comment on upstream bug. |