Am 29.12.2014 um 16:30 schrieb Christian Brunotte: > Am Mon, 29 Dec 2014 16:13:37 +0100 > schrieb Michael Biebl <[email protected]>: > >> control: tags -1 moreinfo >> Am 29.12.2014 um 15:57 schrieb Christian Brunotte: >>> Package: systemd >>> Version: 215-8 >>> >>> Hallo >>> >>> I tried to upgrade from wheezy to jessie but had problems because >>> the install hung several times e.g. when upgrading snmpd and cups. >>> >> >> Are you running systemd as PID 1 while doing the upgrade? > > yes, as you can see on the timestamps, currently (during the upgrade) > PID 1 is still the "old" version, all other related processes have > meanwhile been restarted: > > chammers@sys-251:~$ ps faxu | grep systemd > root 1 0.0 0.0 5776 4284 ? Ss 15:04 0:03 > /lib/systemd/systemd --log-level info --log-target journal-or-kmsg --system > --dump-core --show-status=1 --sysv-console=1 --deserialize 24 > root 379 0.0 0.0 599060 7272 ? Ss 15:04 0:02 > /lib/systemd/systemd-journald > chammers 27185 0.0 0.0 5048 2272 pts/5 S+ 16:16 0:00 \_ grep > systemd > root 3642 0.0 0.0 2916 2108 ? Ss 16:04 0:00 > /sbin/cgmanager -m name=systemd > 103 21560 0.0 0.0 5372 3476 ? Ss 16:10 0:00 > /usr/bin/dbus-daemon --system --address=systemd: --nofork --nopidfile > --systemd-activation > root 4425 0.0 0.0 3748 2716 ? Ss 16:13 0:00 > /lib/systemd/systemd-logind > root 4468 0.0 0.1 20560 10280 ? Ss 16:13 0:00 > /lib/systemd/systemd-journald > >> Which version of systemd did you have installed? > > old: > 2014-12-22 10:54:24 status installed systemd:i386 44-11+deb7u4 > new: > 2014-12-29 16:13:31 status installed systemd:i386 215-8 > >> Can you provide a journal log (ideally, boot with >> systemd.log_level=debug) when doing the dist-upgrade > > I attached so logs so far and enable debugging logs on next boot. > Hopefully I won't encounter the bug after upgrading and rebooting though.
I think I can reproduce the problem in a VM, having wheezy installed and systemd v44 as PID 1. During the dist-upgrade, systemd 215-8 is unpacked, udev 215-8 is unpacked, then udev hangs in the configure/postinst step trying to "systemctl stop udev.service udev-kernel.socket udev-control.socket" at [1] At this point, systemd's postinst is not yet run, and systemd hasn't been re-execd. If I manually run "systemctl daemon-reexec" at this point, the upgrade continues successfully. I remember that I successfully tested dist-upgrades from v44 to v215 but I think the upgrade sequence changed to the recently bumped dependency on udev to "udev (>= 208-8)". The problematic postinst code was added in commit [2]. Looks like everyone running systemd v44 as PID in wheezy will be affected, so I'm inclined to bump the severity to serious. That said, I don't know how to fix this issue, so feedback and comments from the team would be welcome. Michael [1] http://anonscm.debian.org/cgit/pkg-systemd/systemd.git/tree/debian/udev.postinst#n125 [2] http://anonscm.debian.org/cgit/pkg-systemd/systemd.git/commit/?id=0bcf5c01 -- Why is it that all of the instruments seeking intelligent life in the universe are pointed away from Earth?
signature.asc
Description: OpenPGP digital signature
