Martin, I don't have time to break and fix my system again right this instant (I can debug this weekend), but things are definitely unmounted (like /tmp). It occurred close to the closing of the "root slice." I logged in as root, so it could have been a _privileged_ pam-systemd/user systemd issue.
Obviously, / was not unmounted, but a whole bunch of other stuff (like /home, which was not in use) was. I'll check if not logging in as root is a workaround, but it's still a serious bug if logging as root causes systemd to try to unmount everything. As well as getting you a journalctl output. Thanks, Antonio On 10/08/15 23:52, Martin Pitt wrote: > Control: tag -1 moreinfo > > Hello Antonio, > > Antonio Russo [2015-10-08 20:52 -0700]: >> The machine in question has /var, /tmp on ZFS (/tmp is mounted >> "legacy style"). Because these are needed fairly early in the >> boot process, I mount the ZFS filesystems very early in the boot >> process, immediately after systemd-udev-settle. >> >> At the end of the boot, systemd decides to unmount lots of things, >> including the "legacy" mountpoint (/tmp) and zfs-style mountpoint >> /var. Even more strangely, it attempts to unmount / (and fails). > > I can confirm that the *user* systemd instance (not pid 1) attempts to > do that when closing a session, and it looks indeed confusing: > > | systemd[1653]: Reached target Default. > | systemd[1653]: boot.mount: Mount process exited, code=exited status=1 > | systemd[1653]: Failed unmounting /boot. > | systemd[1653]: run-user-1000.mount: Mount process exited, code=exited > status=1 > | systemd[1653]: Failed unmounting /run/user/1000. > | systemd[1653]: dev-dm\x2d2.swap: Unit entered failed state. > | systemd[1653]: sys-kernel-debug.mount: Mount process exited, code=exited > status=1 > | systemd[1653]: Failed unmounting /sys/kernel/debug. > | [...] > | umount[1657]: umount: /boot: umount failed: Operation not permitted > | umount[1658]: umount: /run/user/1000: umount failed: Operation not permitted > | swapoff[1659]: swapoff: Not superuser. > > and so on. This is indeed weird, but should mostly be cosmetical as > the user systemd instance has no privileges to actually do that. > > Do you get the same? Or is it pid 1 in your case, and it actually > unmounts things? Then this would be something different and quite > serious. > > If you see something else, please grab "journalctl -b > > /tmp/journal.txt" (as root) after such a boot, and attach it here. > > Thanks, > > Martin >
