Bugzilla – Bug 134187
NetworkManager segfault if DBUS stopped
Last modified: 2007-06-05 09:57:56 UTC
If you stop DBUS you get a segfault of the Networkmanager. This is IMO not the same problem as if you stop HAL daemon. Because this happen also if do this: * rcnetworkmanager stop * rchal stop * rcnetworkmanager start * rcdbus stop --> segfault I attach a strace ;-)
Created attachment 57650 [details] strace from segfault
The current view of DBUS hackers is that DBUS should not be restarted on a live system. http://mail.gnome.org/archives/networkmanager-list/2005-March/msg00023.html http://mail.gnome.org/archives/networkmanager-list/2005-March/msg00029.html It is a sensible argument. But we can do better than crash, to be sure.
Yes I know this discussion, and I can't agree to the arguments. IMO there must/should be a way to restart DBUS. To reboot after a update is for me a really stupid idea (for me the only acceptable reason for a reboot is a kernel-update or a kernel oops but not a update of a system daemon), but this is not the point in this bug ;)
(In reply to comment #2) > The current view of DBUS hackers is that DBUS should not be restarted on a live > system. DBUS might be linked against libraries that are upgraded by whatever will replace yast online update (say, for security updates). Requiring a reboot just to run against new libraries seems a bit harsh to me.
The plan is for DBUS to be ABI/API compatible once it reaches 1.0, so that won't be a concern.
*** Bug 134179 has been marked as a duplicate of this bug. ***
Alright. I did some work in NM to make it a little more resiliant to DBUS problems. There is still a segfault, however: Now, in libdbus. libdbus itself segfaults when used on a no-longer-valid connection. See backtrace. Reassigning to DBUS maintainer, as the problem is in libdbus. But, as Kay says, "supporting DBUS restarts is like expecting X applications to survive an X restart" ;-)
Created attachment 60126 [details] Backtrace from crash. Segfault is in libdbus.
I'll have a look.
(In reply to comment #7) > But, as Kay says, "supporting DBUS restarts is like expecting X applications > to survive an X restart" ;-) IMO this is complete wrong position. There are many programms which survive a DBUS restart as e.g. HAL, powersave, KPowersave ... This should work. Everything other is a alibi ;-)
This should now be fixed. Right, Timo?
Yes, that's done. Thus, closing as fixed.