[systemd-devel] [PATCH 1/4] Adding halt binary to shutdown the system

[Lennart Poettering]

On Fri, 01.10.10 16:36, Michael Biebl (mbiebl at gmail.com) wrote:

> 
> 2010/10/1 Gustavo Sverzut Barbieri <barbieri at profusion.mobi>:
> > From: Fabiano Fidencio <fidencio at profusion.mobi>
> >
> > This functions are working as follows:
> >    - Send a SIGTERM to all process
> >    - Send a SIGKILL to all process
> >    - Try to umount all mount points
> >    - Try to remount read-only all mount points that can't
> >    be umounted
> 
> What about remote mounts (e.g. NFS requiring portmap) or fuse mounts?
> If you kill their processes before unmounting you can not unmount
> those fs cleanly.

I am pretty sure that both NFS and FUSE mount points can actually be
unmounted properly even if portmap resp. the providing processes are
gone. Since the providing processes will be killed with SIGTERM first,
they should have the chance to get rid of the mount points themselves
safely.

Also note that the code in question is needed only as last resort. At
the point this is executed all mount points and services should already
have been shut down cleanly and in the proper order. This code only
kills and unmounts what is left. Stuff will only be left if processes
actively tried to escape systemd's supervision (by playing games with
/sys/fs/cgroup/systemd) or turned out to be unkillable or suchlike.

In short: I don't think this is a problem and we already have all the
right hooks to get rid of the mount points and services correctly and
cleanly.

Lennart

-- 
Lennart Poettering - Red Hat, Inc.