Bug 1110697
| Summary: | we shouldn't restart logind on package updates, as it kills X11 | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Zdenek Kabelac <zkabelac> |
| Component: | systemd | Assignee: | systemd-maint |
| Status: | CLOSED RAWHIDE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> |
| Severity: | unspecified | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | rawhide | CC: | hdegoede, johannbg, lnykryn, msekleta, s, systemd-maint, vpavlin, xgl-maint, zbyszek, zkabelac |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | Unspecified | ||
| OS: | Unspecified | ||
| Whiteboard: | |||
| Fixed In Version: | systemd-214-2.fc21 | Doc Type: | Bug Fix |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2014-06-23 05:21:38 UTC | Type: | Bug |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
|
Description
Zdenek Kabelac
2014-06-18 09:30:32 UTC
Do you have any logs for this which could indicate why X dies? Are you sure systemd is even involved? What makes you think that? Well the log from my Xsession tells me: (EE) Fatal server error: (EE) systemd-logind disappeared (stopped/restarted?) (EE) (EE) Also - my kernel message log show this warning related to systemd: "kernel does not support evdev-revocation" - could this be relevant ? (I'm using my own build of kernel) I'm pretty sure my Xsession used not to be killed in the past during systemd upgrade. Also I've seen the last processed thin by yum/dnf upgrade to be a systemd related thing (though it's been just short 'view' since the Xsession is killed) Oh, uh. X11 doesn't allow logind going away and coming back? That's quite a limitation. Reassigning to X11 for now. X11 guys, any chance you can make X11 survive if logind goes away and comes back? logind itself is pretty good at keeping state when it does that (it keeps all its state in /run all the time, to make that possible), it would be a pity if we couldn't actually restart logind on upgrades anymore because that kills X11. We could also stop restarting logind on updates, but before I do that, I'd rather like to see if we can't get X11 fixed. (In reply to Lennart Poettering from comment #3) > Oh, uh. X11 doesn't allow logind going away and coming back? That's quite a > limitation. > > Reassigning to X11 for now. X11 guys, any chance you can make X11 survive if > logind goes away and comes back? logind itself is pretty good at keeping > state when it does that (it keeps all its state in /run all the time, to > make that possible), it would be a pity if we couldn't actually restart > logind on upgrades anymore because that kills X11. > > We could also stop restarting logind on updates, but before I do that, I'd > rather like to see if we can't get X11 fixed. This was discussed in detail a while back, and I believe you were included in that discussion, and back then the conclusion was to just not ever restart logind for now. The problem is that logind manages the drm fd for the xserver now, and logind will give us a new fd after a restart, but all our state, all textures, uploaded shaders, etc. are all tied to the old fd. The way I remember it was that there was this master plan to have logind pass its copy of the drm fds to systemd itself (so to pid 1) upon restart, and then when it comes back it would get them back. Once that is implemented in logind making the xserver survive a logind restart is easy. But since that is not implemented yet, for now the only sane thing to do for the xserver if logind goes away is to do is to just die. Hmm, OK, I am too lazy to implement something in logind to keep the original fds around. I figure we should just stop restarting logind from postinst then. Reassigning back. |