At a ~x86 user mode linux I saw today : * Unmounting filesystems * Unmounting /usr/local/portage ... [ ok ] * Unmounting /usr/portage ... [ ok ] * Deactivating swap devices ... [ ok ] /lib/rc/sh/rc-cgroup.sh: line 87: /sys/fs/cgroup/openrc/fsck/tasks: No such file or directory * Removing dm-crypt mappings * crypt-swap ... [ ok ] * Stopping udev ... [ ok ]
This can be easily fixed by workaround that will be applied soon. But lets try to investigate further before: 1). can you describe in you got some special options for mounting /sys? 2). was it your first reboot/shutdown with openrc-0.12, if yes that do you have such error on the next one?
(In reply to Alexander Vershilov from comment #1) > This can be easily fixed by workaround that will be applied soon. > But lets try to investigate further before: > > 1). can you describe in you got some special options for mounting /sys? no - nothing special -it is just an UML > 2). was it your first reboot/shutdown with openrc-0.12, if yes that do you > have such error on the next one? yes - no - vanished now at the 2nd shutdown
(In reply to Toralf Förster from comment #2) > (In reply to Alexander Vershilov from comment #1) > > This can be easily fixed by workaround that will be applied soon. > > But lets try to investigate further before: > > > > 1). can you describe in you got some special options for mounting /sys? > > no - nothing special -it is just an UML > > > 2). was it your first reboot/shutdown with openrc-0.12, if yes that do you > > have such error on the next one? > yes - no - vanished now at the 2nd shutdown ok, thanks so it's a migration problem. I'll look at it more close. Thanks for report. Feel free to report if you'll see that problem once again, if no I'll plan to prepare migration fixes in the next few days.
Does it even make sense for fsck to run inside uml? If not, we can probably add the -uml keyword to it. It sounds like we tried to access a cgroup/tasks file that did not exist when we tried to access it. I believe we check for existence everywhere, but that might be something we should look at.
(In reply to William Hubbs from comment #4) > Does it even make sense for fsck to run inside uml? At least I never did it.
Opened a pull requiest. https://github.com/OpenRC/openrc/pull/8
Hhm, it is still in 0.12.2 - I just upgraded openrc at a stable x86 user mode linux image from 0.11.8 to .12.2 and get now during boot : * Starting rngd ... [ ok ] * Starting sshd ... [ ok ] * Writing a dummy startup record using sadc (see sadc(8))... ... [ ok ] /lib/rc/sh/rc-cgroup.sh: line 87: /sys/fs/cgroup/openrc/vixie-cron/tasks: No such file or directory * Starting vixie-cron ... [ ok ] * Starting local net.core.warnings = 0 [ ok ]
This is added in commit 6697039 and will be part of OpenRC-0.13 and back ported to 0.12.3.