On Monday 29 May 2017 09:31:39 Guilhem Moulin wrote: > On Sun, 28 May 2017 at 23:41:56 +0200, Pali Rohár wrote: > > On Sunday 28 May 2017 21:26:53 Guilhem Moulin wrote: > >> which as I explained in message #86 is due to /etc/init.d/sendsigs > >> killing systemd-udevd at shutdown/reboot time, thereby causing > >> dmsetup to hang. > > > > And removing remaining socket file /run/udev/control before invoking > > cryptsetup fixes hanging. > > The udev regression might affect other shutdown sequences so I don't > think the workaround should be implemented in cryptsetup's init scripts.
Then it is needed to fix broken udev (from systemd source package). > >> The udev/systemd maintainers have fixed #791944, > >> but their fix probably didn't land into Stretch yet. > > > > Seems it is not in Stretch yet: > > http://metadata.ftp-master.debian.org/changelogs/main/s/systemd/systemd_232-23_changelog > > > > Should I create release bug report for Stretch? This seems to be a big > > problem if shutdown/reboot procedure hangs when using encrypted disk. > > Not sure if the fix can be easily backported to 232-xyz, perhaps you > should follow-up on #791944 and ask there first? Asked, sent comment. > Even if the udev > maintainers don't acknowledge the severity they're in a better position > to determine whether the workaround should go to our init scripts or > not. -- Pali Rohár pali.ro...@gmail.com
